Introduction

During the 2023-24 school year, approximately 11% of Michigan’s K-12 students took at least one virtual course. These 154,056 students comprised over 1,019,000 enrollments, 71% of which came from grades 9-12. In addition, 68% of Michigan school districts reported at least one virtual enrollment (Freidhoff et al., 2024). Despite the number of students opting for online learning, virtual pass rates remain lower than non-virtual ones (Freidhoff, 2015). For example, the overall pass rate for virtual courses during the 2023-24 school year was 63% compared to 74% for non-virtual courses (Freidhoff et al., 2024). Perhaps this difference can partially be attributed to the various challenges students face that are unique to online education (Johnson et al., 2023). 

Indeed, the flexibility of learning “anytime, anywhere,” particularly in asynchronous courses, requires students to possess or acquire a strong foundation of self-regulation, metacognitive and time management skills (Johnson et al., 2023; Digital Learning Institute, nd) because students can submit assignments at any time (but typically guided by an end-of-term deadline) and in any order they like, as courses are typically open (as opposed to having release conditions). Certain Michigan Virtual courses, such as those for core subject areas and electives, fall into this category, meaning students may submit any assignment at any time and in any order during the course term. To help students experience success in their online courses, Michigan Virtual provides students with pacing guides against which to benchmark their progress. Pacing guides outline the order in which students should complete course content. In other words, pacing guides show students which assignments they should complete each week to stay on track within their course. 

While there is a growing body of research pointing to the role that the timing of assignment submissions plays in students’ course performance (Carvalho et al., 2022; DeBruler, 2021; Dunlosky et al., 2013; Kim & Seo, 2015; Lim, 2016; Michigan Virtual Learning Research Institute, 2019), limited research exists exploring how the order of students’ assignment submissions is associated with course outcomes. Research that has examined the order of students’ assignment submissions has focused on massive open enrollment courses or university students and, thus, is difficult to generalize to K-12 populations (e.g., Perna et al., 2014; Lim, 2016b). However, Cuccolo & DeBruler (2024) found that submitting assignments out of order was common among students enrolled in Michigan Virtual STEM courses. Submitting assignments out of order seemed to be the rule and not the exception, with about 93% of students submitting at least one assignment out of order. Despite being common, this behavior was not beneficial for students, as students who submitted at least one assignment out of order had final course scores that were 9.5 points lower than students who completely adhered to course pacing guides. In Michigan Virtual STEM courses, the order of students’ assignment submissions is associated with their final course scores (Cuccolo & DeBruler, 2024), suggesting that assignment sequencing may be an understudied aspect of pacing that has important implications for student outcomes. 

Cuccolo & DeBruler’s original study focused solely on STEM courses, as the highly scaffolded nature of these courses was an ideal choice for examining the impact of pacing guide deviations. Because the amount of scaffolding, number and type of assignments, and course structure (linear vs. non-linear) likely vary by content area and individual course, it stands to reason that the extent to which student assignment submission patterns relate to final course scores may differ between subject areas. Preliminary analyses of assignment sequencing patterns in five core subject areas (English Language and Literature, Life and Physical Sciences, Mathematics, Social Sciences and History, and World Languages) revealed that moving out of alignment with course pacing guides was an especially common behavior for students enrolled in World Language courses. As such, the current study aimed to obtain a deeper understanding of the prevalence and impact of this behavior in these World Language courses. 

Methods

Data & Sample Overview

A subset of highly enrolled Michigan Virtual World Languages courses was selected to facilitate generalizability about students’ sequencing behavior within these courses. The table in Appendix A shows the enrollment information used in the selection process and the number of enrollments included in the current study. A Michigan Virtual Technology Integration team member provided enrollment data from the selected courses for Spring 2024 (the most recently completed semester at the time the data was requested and pulled). 

Analysis

Several variables were created to facilitate the analyses and answer the aforementioned research questions. First, because course pacing guides structure assignments sequentially (i.e., Assignments 1, 2, 3, 4, etc.), each student’s assignment submission was benchmarked against the one immediately preceding it—this ‘User-Driven’ variable coded in-sequence assignments with a 0 and out-of-sequence assignments with a 1. The total number of assignments submitted out of order was then calculated for each student. 

Next, to better contextualize students’ deviation from course pacing guides, the number of assignments submitted out of order was divided by the total number of assignments the student submitted and multiplied by 100 (‘Percentage of Assignments Submitted Out of Order’). For example, if a student submitted 50 assignments, 5 of which were out of order, the percentage out of order would be 10 percent. 

The ‘Magnitude’ variable represents the difference between the intended submission order of consecutively submitted assignments and was used to understand the degree to which students submitted assignments out of alignment with the course pacing guides. For example, if a student submitted assignment 5 and then 17, the magnitude value would be 12. Each student’s respective magnitude values for all submitted assignments were averaged (‘Average Magnitude’) to obtain a value that captured, on average, the extent to which they deviated from course pacing guide expectations. 

Finally, based on how students’ assignment submission patterns were categorized, students were assigned to one of two user sequence groups. Students who submitted at least one assignment out of order were assigned to the out-of-sequence group. In contrast, if students submitted all their assignments in order (aligned to pacing guide expectations), they were assigned to the in-sequence group. Appendix A contains the names and definitions of key variables referenced in the report and examples when applicable. Table 1 below provides an example of the data layout for readers. 

Results

Sample Description

After removing duplicate enrollments (students who were enrolled in more than one of the courses selected for the study) and outliers (students who had completed fewer than 50% of their assignments to ensure reliability, given analyses focused on understanding how assignment submission patterns related to course performance), the final sample consisted of 1,873 students. On average, students in the sample had completed about 3.29 (SD = 2.99) online courses. During the semester in which the data were collected (Spring 2024), students carried a Michigan Virtual (MV) online course load of about one course (SD = 0.73). Most students (60.86%, n = 1140) attended schools where the Non-White School Population was 25% or less. Additionally, about 39.94% of students came from Mid-Low Poverty schools, whereas only 4.91% attended high-poverty schools.

When examining the sample by school locale classification, approximately 25.79% (n = 483) of students were from large suburban areas, followed by 13.61% (n = 255) from rural fringe areas. Regarding entity type, most students (86.92%, n =  1628) came from LEA (Local Education Agency) Schools. American Sign Language 1B had the highest student enrollment of the courses included in the study. Please review the table in Appendix B for a breakdown of the number of students in each course. 

Research Questions

What does students’ assignment sequencing look like in World Language courses?

What percentage of students go out of sequence? Is going out of sequence a common behavior?

The majority of students sampled went out of sequence at least once (n = 1,817, 97.01%). This left only 56 students (2.99%) who completed their World Language course fully aligned with their course pacing guides. As such, going out of sequence appeared to be the norm, implying that students in World Language courses are more likely to deviate from their course pacing guides than adhere to them. 

What are the average and median number of assignments submitted out of sequence?

The total number of assignments submitted out of order ranged from zero (students fully adhered to course pacing guides) to 90. The average number of assignments submitted out of order was approximately 31 (SD = 19.90). The median (a metric referring to the middle value of a data set when organized in ascending order) number of assignments submitted out of order was 29, meaning half of the students sampled submitted fewer than 29 assignments out of order. In contrast, half submitted more than 29 assignments out of alignment with course pacing guides.

What is the average and median percentage of completed assignments submitted out of sequence? 

To put the number of assignments students submitted out of sequence in the context of the number of available course assignments, the percentage of assignments submitted out of sequence was examined. This represents the number of assignments submitted out of sequence divided by the total number of assignments completed and multiplied by 100. While the percentage of assignments submitted out of sequence ranged from zero to 97.70%, the average was 44.70%. The median was slightly higher, with half of the students submitting more than 47% of assignments out of sequence and half submitting fewer than that.

What is the average and median magnitude of assignments submitted out of sequence?

In addition to examining the number of assignments submitted out of sequence, the extent to which assignments were out of alignment with course pacing guides was analyzed. On average, students were about three and a half assignments “off” from the intended pacing guide order (SD = 2.96). This may look like, for example, a student submitting assignment eight when the course pacing guide recommended submitting assignment eleven. While some students moved through the course completely aligned with pacing guide expectations, others were as much as approximately 15 assignments “off.” Please refer to Table 2 for descriptive information about key study variables. 

What does the relationship between students’ assignment sequencing and course performance look like in World Language courses?

Correlations helped describe the relationship between students’ assignment sequencing (percentage of assignments submitted out of order, average magnitude) and students’ course performance (final course scores). Correlations show the changes in one variable relative to changes in the other. 

The negative correlation observed between final course scores and the percentage of assignments submitted out of order means that as one variable decreased, the other increased. For example, as students’ final course scores decreased, the percentage of assignments submitted out of order increased. Likewise, final course scores increased as the percentage of assignments submitted out of order decreased. A similar relationship was observed between the extent to which assignments were submitted out of order (average magnitude) and final course scores. As students’ magnitudes increased, scores decreased; as magnitude values decreased, final scores increased. A correlation matrix is provided in Appendix C for those interested in examining the specific correlation coefficients. 

In short, sequencing variables and final course scores moved in opposite directions. An important caveat to these findings is that correlations only describe the relationship between variables and do not isolate the cause of the relationship (i.e., it cannot be said which variable causes the observed relationship).

What does student performance look like at each quartile of the percentage of assignments submitted out of order?

To understand how out-of-sequence movement related to course performance, students’ average final course scores at each quartile¹ of the percentage of assignments submitted out of sequence were examined.

As illustrated in Table 3, students in the 1st quartile—those who completed the smallest percentage of assignments out of order—achieved the highest average final course scores (M = 88.3). In contrast, students in the 4th quartile, who submitted the largest percentage of assignments out of sequence, earned the lowest average scores (M = 78.7). This 9.6-point difference between the highest and lowest quartiles equals roughly one letter grade. Notably, the most significant drop in scores occurred between the 1st and 2nd quartiles, where average scores declined by 5.8 points, suggesting that even moderate increases in out-of-order submission behavior may be linked to a meaningful decline in academic performance.

What does student performance look like at each quartile of the average magnitude of assignments submitted out of order?

A similar trend emerged when analyzing the magnitude of out-of-order assignment submissions. Students in the 1st quartile—those with the lowest magnitude values—achieved the highest average final course scores (M = 87.3). In contrast, those in the 4th quartile, representing students with the highest magnitude values, had the lowest average scores (M = 78.2). This 9.1-point gap between the first and fourth quartiles is nearly equivalent to a full letter grade. The most pronounced drop occurred between the 2nd and 3rd quartiles, where average final course scores declined by 5.1 points, indicating that even modest increases in the degree of out-of-order submission may be associated with noticeable decreases in academic performance.

Reviewing Table 3 can provide additional information about the average final course scores at each quartile of the respective sequencing variables. A similar pattern was observed when looking at median final course scores at each quartile of the percentage of assignments submitted out of order and average magnitude; the reader can review these trends in Table 4.

Discussion

The current study highlights the prevalence of submitting assignments out of alignment with course pacing guides in Michigan Virtual self-paced World Language courses. Nearly all students in the sample (97%) deviated from course pacing guides at least once, and on average, submitted approximately 44% of their completed course assignments out of order. These findings exceed those reported in prior studies. For example, Cuccolo & DeBruler (2024) found that 93% of students in Michigan Virtual online STEM courses submitted at least one assignment out of order, with an average of approximately 38% of assignments submitted out of sequence. Additionally, a brief visual inspection of the World Language course data supports anecdotal evidence from instructors: a substantial number of students fail to submit assignments requiring video recordings or scheduled meetings with instructors, assignment types that are especially critical in these particular courses for formative skill assessment and feedback. Taken together, these findings suggest that students in World Language courses may be particularly prone to deviating from the recommended assignments sequence, underscoring the importance of monitoring students’ assignment submission patterns in this context. 

Consistent with Cuccolo & DeBruler’s (2024) findings, this study also observed a negative relationship between moving out of alignment with course pacing guides and final course scores. Specifically, final course scores steadily declined as students submitted a greater percentage of assignments out of order, and strayed further from the intended assignment sequence (i.e., higher magnitude). The most significant decline in scores occurred between the 1st and 2nd quartiles for the percentage of assignments submitted out of order, and between the 2nd and 3rd quartiles for average magnitude. Practically speaking, this suggests critical thresholds: a noticeable decline in student grades may be seen as they begin to submit more than about a quarter of assignments out of order, or when they are more than one assignment “off” from pacing guide recommendations. As such, these thresholds may serve as useful benchmarks and noteworthy intervention points for teachers and mentors. 

Overall, the findings suggest that both the frequency and degree of out-of-order assignment completion are meaningfully associated with student performance. Students who submitted fewer assignments out of order and stayed closer to the intended sequence (i.e., lower magnitude) earned notably higher final course scores. As students increased the proportion or the extent to which they deviated from the expected assignment order, their average course scores declined by nearly a full letter grade between the lowest and highest quartiles for both predictors. These patterns highlight the potential academic consequences of straying from the designed learning sequence in online courses. As such, teachers and mentors should be mindful of intervening early when monitoring student performance. 

While the current study’s findings provide actionable insights into how assignment submission patterns are associated with course scores, it is essential to note that this study was correlational by design. In other words, the causal agent in the observed relationship between assignment sequencing and final course score is unclear. Other unstudied factors may contribute to the observed relationship. For example, moving out of alignment with course pacing guides may be part of a broader pattern of student behaviors related to performance. For instance, self-regulatory and metacognitive skills have been positively associated with student achievement (Xu et al., 2023). These skills encompass students’ ability to engage with a task cognitively, reflect and evaluate their learning, manage resources, and direct their efforts (Xu et al., 2023). Self-regulatory and metacognitive skills are increasingly important in an online learning environment, where students have increased autonomy over when, where, and how they engage with course content (Xu et al., 2023). Without a strong foundation, students may struggle to engage deeply with assignments and feedback, scaffold knowledge, and manage their time appropriately. Thus, self-regulated learning could be effective in better understanding the relationship between assignment submission patterns and student performance. 

To promote adherence to course pacing guides and foster effective course navigation, teachers and mentors should incorporate transparent and proactive communication practices into their routines. In particular, setting course expectations early by helping students understand course structure, workload, pacing, and tips for success can help students prepare for the demands of self-paced online learning (Cuccolo & Green, 2024). Monitoring the gradebook and benchmarking student progress against course pacing guides is also recommended. The thresholds identified in this study (e.g., exceeding about a quarter of out-of-sequence submissions or being more than one assignment “off” from the pacing guide’s intended submission order) can serve as timely indicators for intervention. Personalized feedback remains a highly effective strategy for engaging students in online learning as it works double duty as both a relationship-building strategy and a way to monitor/motivate academic progress (DeBruler & Harrington, 2024). Regular communication between mentors, teachers, and their students helps bridge information gaps and close feedback loops. Additionally, mentors can also help contextualize student behavior for teachers (Cuccolo & DeBruler, 2023). While it may be unrealistic to believe students will complete all assignments in order when enrolled in a self-paced course, teachers and mentors can help encourage pacing behaviors that set students up for success. 

Appendix

References

Cuccolo, K., & Green, C. (2024). Starting Strong: Understanding Teacher-Student Communication in Online Courses. Michigan Virtual. https://michiganvirtual.org/research/publications/understanding-teacher-student-communication

Cuccolo, K. & DeBruler, K. (2023). Examining Mentors’ Navigation of Online Environments and Use of Student Support Practices. Michigan Virtual. https://michiganvirtual.org/research/publications/examining-mentors-navigation-of-online-environments

DeBruler, K. (2021). Research On K-12 Online Best Practices. Michigan Virtual. 

DeBruler, K. & Harrington, C. (2024). Key Strategies for Supporting Disengaged and Struggling Students in Virtual Learning Environments. Michigan Virtual. https://michiganvirtual.org/research/publications/key-strategies-for-supporting-disengaged-and-struggling-students-in-virtual-learning-environments/

Digital Learning Institute. (n.d.). “What is Self Paced Learning? Definition, Benefits and Tips.” [Blog]. Retrieved from https://www.digitallearninginstitute.com/blog/what-is-self-paced-learning-definition-benefits-and-tips

Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques: Promising directions from cognitive and educational psychology. Psychological Science in the Public Interest, 14(1), 4-58. https://doi.org/10.1177/1529100612453266

Freidhoff, J. R. (2015). Michigan’s K-12 virtual learning effectiveness report 2013-14. Lansing, MI: Michigan Virtual University. Retrieved from http://media.mivu.org/institute/pdf/er_2014.pdf

Freidhoff, J. R., DeBruler, K., Cuccolo, K., & Green, C. (2024). Michigan’s K-12 virtual learning effectiveness report 2022-23. Michigan Virtual. https://michiganvirtual.org/research/publications/michigans-k-12-virtual-learning-effectiveness-report-2022-23/

Johnson, C. C., Walton, J. B., Strickler, L., & Elliott, J. B. (2023). Online teaching in K-12 education in the United States: A systematic review. Review of Educational Research, 93(3), 353-411. https://doi.org/10.3102/00346543221105550

Kim, K. R., & Seo, E. H. (2015). The relationship between procrastination and academic performance: A meta-analysis. Personality and Individual Differences, 82, 26-33. https://doi.org/10.1016/j.paid.2015.02.038

Lim, J. M. (2016). Predicting successful completion using student delay indicators in undergraduate self-paced online courses. Distance Education, 37(3), 317-332. https://doi.org/10.1080/01587919.2016.1233050

Lim, J. M. (2016b). The relationship between successful completion and sequential movement in self-paced distance courses. International Review of Research in Open and Distributed Learning, 17(1), 159-179. https://doi.org/10.19173/irrodl.v17i1.2167

Michigan Virtual Learning Research Institute. (2019). Pacing Guide For Success In Online Mathematics Courses. https://michiganvirtual.org/blog/pacing-guide-for-success-in-online-mathematics-courses/

Perna, L. W., Ruby, A., Boruch, R. F., Wang, N., Scull, J., Ahmad, S., & Evans, C. (2014). Moving through MOOCs: Understanding the progression of users in massive open online courses. Educational Researcher, 43(9), 421-432. https://doi.org/10.3102/0013189X14562423

Xu, Z., Zhao, Y., Zhang, B., Liew, J., & Kogut, A. (2023). A meta-analysis of the efficacy of self-regulated learning interventions on academic achievement in online and blended environments in K-12 and higher education. Behaviour & Information Technology, 42(16), 2911-2931. https://doi.org/10.1080/0144929X.2022.2151935

Introduction

This report presents an analysis of information on virtual learners reported by schools to the state and shares findings in a highly consumable way to aid the evaluation of virtual learning programs. This year’s report is the 12th edition of this annual publication and completes 14 years of data on K-12 virtual learning in Michigan.

The report is organized into several sections. Each section is meant to capture the essential findings without being overly intensive; however, data tables have been included in the appendices to provide those interested with more in-depth information. Information about the report’s methodology is captured in Appendix A. Please note that in some tables and figures, percentages may not sum to 100% due to rounding.

Schools

Number of Districts and Schools

For the 2023-24 school year, 613 districts reported having at least one virtual enrollment. This represented approximately 68% of the 899 Michigan public school districts for the year. See the MI School Data Report for a breakdown of the district count. Within those districts, 1,421 schools reported virtual enrollments, 54 fewer than the prior year. When looking over the last two years, schools fell into three categories, which are also captured in Table B1:

• Leaving – 198 schools had virtual enrollments the prior year (2022-23) but did not report any virtual enrollments in 2023-2024. Last year, those schools accounted for a total of 27,830 virtual enrollments and had a pass rate of 70%.
• Returning – 1,277 or 90% of schools in this year’s dataset reported virtual enrollments in both 2022-23 and 2023-2024. This year, these schools generated 991,942 enrollments and had a pass rate of 63%, which was three percentage points lower than their rate in 2022-23.
• New – 144 schools reported virtual enrollments this year that did not last year. Those schools accounted for 27,719 enrollments, with a pass rate of 77%.

394,331 of this year’s enrollments came from 30 schools that reported 1,000 or more enrollments than they did in 2022-23. On the other hand, 27 schools reported decreases of 1,000 or more virtual enrollments this year. Despite these declines, these schools yielded 65,155 virtual enrollments this year. See Table B2. These declines are less drastic than in previous years, suggesting that volatility from the pandemic may be starting to slow. About 17.4% of schools in both years saw their pass rates increase by 10 or more percentage points from the prior year. See Table B3.

By Grade Level

There were 1,019,661 virtual enrollments across the 1,421 schools. Students in 12th grade generated the most virtual enrollments (249,246), representing 24% of all virtual enrollments. There continued to be a smaller percentage of high school virtual enrollments than before the pandemic. In the 2019-20 school year, 81% of the virtual enrollments came from students in high school; this year, high school enrollments accounted for approximately 71% of the virtual enrollments. It seems likely that this percentage will continue moving upward over the next several years.

The overall pass rate for virtual enrollments was 63%, a decrease of 2 percentage points over the prior year, which may be expected as high school enrollments continue to rebound. See Table G1 for a more specific breakdown of all the completion statuses. This ranged from a high of 82% in first grade to a low of 45% in 9th grade.

This year, elementary grades tended to see larger percentage decreases than last year (two to eight percentage points), whereas the middle school grades (6^th, 7^th, 8^th) saw increases of between one to four percentage points. Among the high school grades, 9^th and 10^th grades saw small decreases (three and two percentage points, respectively), while 11th and 12th grades saw small increases (one and two percentage points, respectively). See Table B4 for more information.

The fairly consistent pattern of a higher pass rate in non-virtual coursework continued. For 2023-24, virtual learners had a 63% pass rate in their virtual courses but a 74% pass rate for their non-virtual coursework. See Table B5. As a pre-pandemic comparison, the 2019-20 school year virtual pass rate was 12 percentage points lower than those students’ non-virtual pass rate.

By School-Level Virtual Pass Rate

Of the 1,421 schools with virtual enrollments, 411, or 29%, had school-level virtual pass rates of 90% to 100%. This was one percentage point higher than the prior year. Approximately 61% percent of the schools (865) had virtual pass rates of 70% or higher. This was three percentage points higher than the prior year. See Table B6. Thus, even though the overall pass rate in the state dropped year over year, a higher percentage of schools experienced high levels of student performance.

By Entity Type

LEA schools and PSA schools accounted for almost all the virtual enrollments, with 60% and 38%, respectively. Virtual enrollments came from 1,251 (88%) LEA schools, while only 128 (9%) of the schools were PSAs. See Table B7. LEA schools had a higher pass rate (64%) than PSA schools (60%), continuing last year’s trend, although pass rates for both entities dropped. See Table B8 or, for a more in-depth look at the completion statuses, see Table G2.

To put this in perspective, as of January 2025, roughly 44% of all LEA schools had a virtual program, compared to approximately 23% of ISD schools. The percentages of State and PSA schools with virtual programs were similar at approximately 33 and 35%, respectively.

By Full-Time Virtual Schools

The number of full-time virtual schools (76) decreased by one from the prior year. Fifty-five of the 76 full-time virtual schools (72%) were LEA schools. PSA schools (18) accounted for 24% of the full-time virtual schools. See Table B9. Despite the sizable difference in the number of schools, PSAs reported more virtual enrollments (62%) from full-time virtual students statewide compared to LEAs (37%). PSA full-time virtual learners saw higher virtual pass rates (62%) than their counterparts in LEA schools (55%). See Table B10 and Table G3. Overall, the number of virtual enrollments from full-time virtual schools increased from 449,188 in 2022-2023 to 499,793 this year. Approximately 49% percent of the virtual enrollments came from full-time virtual learners.

A quick note about full-time virtual schools: Historically, full-time virtual schools have only provided students with 100% of their learning online. Thus, it was safe to designate all enrollments from such a school as being part of a full-time virtual program. Over the last several years, however, LEAs have started to add full-time virtual options to their offerings. In some cases, this is a separate school, which makes it analogous to cyber schools. However, it seems that schools are increasingly offering multiple forms of online learning (“Full Virtual,” “Face Virtual,” and “Supplemental Virtual”) from the same building code. See page 15 of the Educational Entity Master Glossary for more information on these field values. This means that some schools report various forms of virtual (and sometimes non-virtual) learning from a single building code. Case in point, 12% of the enrollments from virtual learners in LEA full-time programs were not flagged as being delivered virtually, indicating what may be more of a hybrid approach.

By Part-Time Virtual Schools

About 95% of the schools offering virtual learning do so to supplement their face-to-face course offerings. These 1,398 schools, referred to in this report as part-time virtual schools, were predominantly LEA schools (89%). See Table B11. Eighty-eight percent of the part-time virtual students were enrolled through LEA schools, and 11% through PSA schools. LEA schools accounted for 431,372 virtual enrollments or 83% of the part-time enrollments. In total, enrollments from part-time virtual schools accounted for approximately 51% of all the virtual enrollments for the year. LEA schools had a pass rate of 68%, whereas PSA schools had a pass rate of 54%. Overall, the pass rate for the part-time virtual schools (67%) was seven percentage points higher than the rate for the full-time virtual schools (60%). See Table B12 and Table G4.

By School Emphasis

Seventy-nine percent of schools with virtual learning were designated as General Education and produced 608,000 (60%) virtual enrollments. Schools with Alternative Education as their emphasis accounted for 409,964 (40%) of the virtual enrollments. See Table B13. There was a considerable difference in virtual pass rates between these two types of schools. General Education schools had a 72% virtual pass rate, whereas Alternative Education schools had a 51% virtual pass rate (see Table B14 and Table G5), though this varied by entity type. LEA schools, for instance, had a 76% virtual pass rate for General Education schools and a 53% virtual pass rate for Alternative Education schools. See Table B15.

By Number of Virtual Enrollments

Fifty-six percent of schools with virtual enrollments had 100 or more virtual enrollments. These schools were responsible for 98% of the virtual enrollments (1,000,507). See Table B16.

Interestingly, last year’s trend that schools with fewer virtual enrollments per student performed better was absent. The trend in which, in general, schools with 1-2 enrollments per student or 5+ enrollments per student performed better than those with 3-4 enrollments per student continued. Indeed, 40% of schools with an average of one to two virtual enrollments per virtual learner had a pass rate of 90-100%, and 40% of schools with an average of five or more virtual courses per student also had a pass rate between 90-100%. See Table B17.

By Locale

Suburban schools represented 36% of schools with virtual enrollments. Rural settings provided the second most schools with 34%. Suburban schools also tallied the largest percentage of the virtual enrollments at 38%. Rural schools were the next closest, providing 30% of the enrollments. See Table B18. In each of the four locales, schools with 100 or more virtual enrollments accounted for the largest percentage of schools. See Table B19. Virtual pass rates varied by locale. City schools, which had the highest virtual pass rate in 2022-23, had the lowest pass rate this year at 62%. This represents an eleven-percentage point drop. Suburban schools had the highest at 68%. See Table B20. On the other hand, Rural and Suburban schools had 47% of their schools achieve building-wide virtual pass rates of 80% or higher. See Table B21. For more information about locales, including definitions, please see pages 23-24 of the Educational Entity Master Glossary.

By School Free or Reduced-Price Lunch Categories

Schools were categorized into one of four categories based on the percentage of all learners at the school (not just virtual learners) that qualified for free or reduced-price (FRL) meals:

• Low FRL (<=25%)
• Mid-Low FRL (>25% to <=50%)
• Mid-High FRL (>50% to <=75%)
• High FRL (>75%)

Similarly to last year, none of the categories had 50% or more of its schools report virtual learners. Mid-high FRL had the highest percentage at 47%. The higher numbers of schools from each category with virtual learners may be attributable to the pandemic, and numbers will likely continue to stabilize in the coming years. See Table B22.

While High FRL schools represented only 37% of schools with virtual learners (369), they accounted for 44% of the virtual enrollments. Mid-High FRL schools accounted for 36% of the enrollments. Low FRL schools, on the other hand, reported less than 5% of the virtual enrollments. The virtual pass rate for Low FRL schools was 85% compared to 57% for Mid-High FRL and 60% for High FRL schools. See Table B23.

By High Enrollment Schools

There were over 200 schools that had 1,000 or more virtual enrollments. It is worth noting that these high enrollment schools were less defined by having a larger number of virtual students (88,136 versus 67,835) and more about the number of virtual enrolls per student (8.9 virtual courses per student in the high enrollment schools compared to 3.5 in the non-high enrollment schools).

For high enrollment schools, both LEAs (153 schools) and PSAs (46 schools) had virtual pass rates of 61%. The corresponding virtual pass rates for virtual enrollments from schools with less than 1,000 enrollments moved in opposite directions. For non-high enrollment schools, the LEA schools’ virtual pass rate was nine percentage points higher (70%), whereas the PSA schools’ virtual pass rate was seven percentage points lower (54%). See Table B24 and Table B25.

There was close to a 50-50 split in the number of Alternative Education (104) and General Education (100) high enrollment schools. The virtual pass rate for these schools was 50% and 70%, respectively. In both cases, students from non-high enrollment schools performed better – three percentage points and five percentage points, respectively. See Table B26 and Table B27.

Courses

Number of Courses

The 1,019,661 virtual enrollments came from 1,091 different course titles, as determined by unique SCED codes.

Courses by Subject Area

English Language and Literature was, again, the subject area with the highest number of virtual enrollments (186,987)—18% of all virtual enrollments. In fact, since 2013-14, English Language and Literature has consistently been the highest enrollment subject area. Mathematics (17%), Social Sciences and History (16%), and Life and Physical Sciences (15%) were the next highest enrollment subject areas. In high enrollment subject areas (greater than 75,000 virtual enrollments), virtual pass rates varied from a low of 59% in Mathematics to a high of 64% in both Physical, Health, and Safety Education and Social Sciences and History. See Table C1 and Table G6. Six of the 23 subject areas (Agriculture, Food, and Natural Resources; Engineering and Technology; Manufacturing; Miscellaneous; Nonsubject Specific; and Transportation, Distribution, and Logistics) had virtual pass rates that were equal to or greater than the non-virtual pass rates for these students. See Table C2. For comparison, in the years just prior to the pandemic, only one or two of the subject areas saw equal or better performance in the virtual courses.

Highest Virtual Enrollment Courses

For English Language and Literature, the most highly enrolled in virtual courses were 9th, 10th, 11th, and 12th grade English/Language Arts. Of those four, the pass rate was lowest for 9th grade English/Language Arts (46%) and rose fairly consistently for each subsequent grade level to finish at 63% for 12th grade English/Language Arts. The next three course titles were at the 6-8 grade level, followed by a multi-grade course and two in the K-5 level (grade 1 and kindergarten). See Table C3.

Geometry, Algebra I, and Algebra II had the most Mathematics enrollments, each having over 27,000. Middle school Mathematics courses ranged from 6,900 to 8,800 enrollments. The pass rate across the top 10 most enrolled-in virtual Mathematics courses ranged from a low of 46% for Algebra I to a high of 77% in Consumer Mathematics. See Table C4.

Biology (34,536), Chemistry (20,687), and Earth Science (14,904) were the highest enrollment course titles, responsible for 10% or more of the virtual enrollments in Life and Physical Sciences courses. Three others—Physical Science, Earth and Space Science, and Environmental Science—each had more than 7,000 enrollments. Earth Science had the lowest pass rate (52%) of those in the top 10; the highest was 73% in both Science (grade 6) and Science (grade 7). See Table C5.

For Social Sciences and History, both U.S. History–Comprehensive (22,878) and World History and Geography (19,245) yielded 10% or more of the virtual enrollments. Three other titles had more than 10,000 enrollments (Economics, World History—Overview, and U.S. Government—Comprehensive). Pass rates for the top 10 most enrolled in courses ranged from a low of 49% in World History and Geography to a high of 77% in Psychology. See Table C6.

Thirty-three AP courses were taken virtually in 2023-24. There were just under 4,900 virtual AP enrollments, up from around 4,000 enrollments the prior year. AP Psychology was the most popular course, accounting for 17% of the enrollments. The pass rate for AP courses taken virtually was 87%. See Table C7. The pass rate for non-virtual AP courses taken by virtual learners was 95%.

There were just over 2,600 students who took at least one AP course virtually. For close to half (45%) of these students, AP courses were the only virtual courses taken. Students whose only virtual courses were AP accounted for over 2,200 enrollments or about 45% of the virtual AP enrollments for the year. These students had a virtual pass rate of 91%, four percentage points higher than the state virtual AP pass rate. In addition, of the 925 schools that had virtual enrollments in 9^th, 10^th, 11^th, or 12^th grade, 269 schools (29%) had virtual AP enrollments.

Subject Area Enrollments by Locale

Course enrollment patterns were quite consistent across locales. For instance, in Social Sciences and History, Life and Physical Sciences, and Physical, Health, and Safety Education, the difference in the percentage of virtual enrollments across the locales (Rural, Town, Suburb, and City) was within one to two percentage points. See Table C8. However, pass rates in virtual courses varied across subject areas and locale. For instance, in English Language and Literature, the pass rates ranged from City and Rural at 58% and 60%, respectively, up to 66% for Suburban schools. These trends—City and Rural schools lagging behind the performance of students in other locales and Suburban schools outperforming students in other locales—were also true for the other core subjects of Mathematics, Life and Physical Sciences, and Social Sciences and History. See Table C9. Last year, it was Town and Rural schools that followed the lowest-performing locale trend.

Subject Area Enrollments by Student Sex

Males and females enrolled in subject areas in similar proportions. In the four highest enrollment subject areas (English Language and Literature, Mathematics, Life and Physical Sciences, and Social Sciences and History), the proportion of enrollments from males and females was the very same or within one percent of the other. Pass rates, however, showed more variability by student sex. In 15 of the 19 subject areas with reported pass rates for both sexes, females outperformed males—a trend that has been consistent in past years. Overall, females had a 65% virtual pass rate whereas males had a 62% pass rate. See Table C10.

Courses by Virtual Method

Schools classified virtual courses into one of three methods: Blended Learning, Digital Learning, or Online Learning. See pages 346 and 347 of the Michigan Student Data System Collection Details Manual Version 4.0.

• Blended Learning – A hybrid instructional delivery model where pupils are provided content, instruction, and assessment at a supervised educational facility where the pupil and teacher are in the same physical location and in part through internet-connected learning environments with some degree of pupil control over time, location, and pace of instruction. For a course to be considered blended, at least 30% of the course content is delivered online.
• Digital Learning – A course of study that is capable of generating a credit or a grade that is provided in an interactive internet-connected learning environment that does not contain an instructor within the online environment itself. There may be a teacher of record assigned to the course, but this teacher does not provide instruction to students through the online environment. For a course to be considered online as opposed to blended, all (or almost all) the course content is delivered online.
• Online Course – A course of study that is capable of generating a credit or a grade that is provided in an interactive internet-connected learning environment, where pupils are separated from their teachers by time, location, or both. For a course to be considered online as opposed to blended, all (or almost all) the course content is delivered online.

Blended Learning enrollments accounted for 7% of the virtual enrollments and had a pass rate of 68%. Digital Learning totaled 9% of the enrollments with a 63% pass rate. Online courses represented most of the enrollments (84%) and yielded a pass rate of 63%. Perhaps worth noting are the striking 1,273 schools (128,675 students) with online course enrollments as compared to the 188 schools (11,590 students) with blended learning enrollments. See Table C11.

Students

By Grade Level

For the 2023-24 school year, 154,087 Michigan K-12 students, approximately 11% of students in the state, took at least one virtual course. This change represents approximately a 3% decrease from the previous year, and a 26% decrease from 2021-22. Seventy-five percent of virtual learners came from the high school grades. Each elementary and middle school grade level tended to be around 2% to 5% of the virtual learners with each of the high school grade levels between 13% to 27%. See Table D1.

By Student Sex

There were slightly more females (78,908) enrolled in virtual courses than males (75,209), though from a percentage perspective, each represented about half of the population. Females had a 3% higher pass rate (65% compared to males at 62%), continuing the trend seen in past years of females outperforming their male counterparts on this measure. See Table D2 and Table G7.

By Race/Ethnicity

White students represented 62% of virtual students with African American or Black students totaling the second highest percentage with 19%. While African American or Black students represented 19% of virtual students, approximately 60% of schools enrolled at least one African American or Black student (93% of schools enrolled at least one White student). Asian students had the highest pass rate at 80%. See Table D3 and Table G8. These demographics are similar to the statewide K-12 demographics for 2023-24.See Student Enrollment Count Report.

By Poverty Status

Sixty-four percent of virtual learners were classified as living in poverty. This is 1% higher than the prior year and approximately 10 percentage points higher than the percentage of K-12 students statewide who were economically disadvantaged. See Student Enrollment Count Report. Students living in poverty took 71% of the virtual enrollments for the year. When looking from a school-level perspective, 93% of Michigan schools enrolled at least one student in poverty in a virtual course. Comparatively, 86% of schools enrolled at least one student who was not in poverty in a virtual course. The pass rate for students in poverty (58%) was 19 percentage points lower than students who were not in poverty (77%). See Table D4 and Table G9. In 2022-23, the performance gap was 17 percentage points.

Prior to the pandemic, the data consistently showed that students in poverty performed better in their non-virtual courses. The 2020-21 and 2021-22 school years deviated from that pattern. In 2021-22, we saw that students in poverty had a higher pass rate in their virtual courses (64%) than they did in their non-virtual courses (62%). For the 2022-23 year, this trend was reversed and students in poverty did better in their non-virtual courses (64% compared to 60%). This remained consistent for the 2023-24 year with both students in poverty and those not performing better in their non-virtual courses. See Table D5.

Seventy-four percent of full-time virtual learners were in poverty compared to 60% of part-time virtual learners. The pass rate for full-time virtual learners in poverty was 56% compared to 60% for part-time virtual learners. See Table D6.

To get a sense of how the poverty level of schools might impact virtual learning patterns, we categorized schools into one of four categories based on the percentage of all learners at the school (not just virtual learners) that qualified for free or reduced-price (FRL) meals:

• Low FRL (<=25%)
• Mid-Low FRL (>25% to <=50%)
• Mid-High FRL (>50% to <=75%)
• High FRL (>75%)

About 6% of all Michigan K-12 students who attended Low FRL schools were virtual learners. Nine percent of the state’s students in Mid-Low FRL, and 12% of those in Mid-High FRL schools were virtual learners. Seventeen percent of students in High FRL schools took virtual courses in the 2023-2024 school year. See Table D7. Although overall virtual enrollments have steadily decreased since pandemic highs, this trend has remained relatively stable. From 2020-21 through the current year, schools with higher percentages of students qualifying for FRL also saw higher percentages of virtual learners.

By Special Education Status

Students using special education services made up 13% of the virtual learners and 14% of the virtual enrollments. These percentages are similar to the statewide percentage of students using special education services (14%) for the 2023-24 school year. See the Student Enrollment Counts Report. Seventy-six percent of Michigan schools enrolled a Special Education student in a virtual course. Students using special education services had a virtual pass rate of 56% compared to 64% for those who did not. See Table D8 and Table G10.

Table D9, shows how virtual enrollments varied by a students’ primary disability. Just over 7,900 students had “Specific Learning Disability” listed as their primary disability. This translated to 39% of the virtual learners receiving special education services. The second and third largest groups were students with Other Health Impairments (4,451) and Emotional Impairment (2,667). These groups represented 22% and 13% respectively of virtual learners receiving special education services. Students with Physical Impairment had the highest virtual pass rate at 79%.

Table D10 shows how the percentage of virtual learners using special education services by primary disability compares to the overall state rates. For instance, only about five percent of the states’ students with an IEP have “Emotional Impairment” listed as their primary disability. However, 26% of those students ended up taking at least one virtual course in 2023-24. These two tables can assist in tracking how virtual learning is being used to target specific disabilities and how well performance follows.

By Home-School / Nonpublic Student Status

Table D11 shows virtual learning data for home-schooled and nonpublic students enrolling in a public school to augment their education. There were just over 7,500 such students, and this group of students generated over 34,000 virtual enrollments, an increase of approximately 14,000 enrollments from the 2022-23 year. These students had a 93% virtual pass rate.

By Full-Time or Part-Time

Thirty-three percent of students (50,803) were enrolled in cyber or full-time virtual schools. Students in these schools accounted for 449,793 or 49% of the virtual enrollments for the year. The pass rate for full-time virtual students was 60%. Sixty-eight percent of virtual learning students were part-time virtual learners, taking some courses virtually to supplement their face-to-face schedule. This subset made up 51% of the virtual enrollments and had a pass rate of 67%. See Table D12. The 67% virtual pass rate was six percentage points lower than the non-virtual pass rate for these students. See Table D13.

Another way to conceptualize full/part-time status is to look at the percentage of a student’s enrollments that were delivered virtually. There were many students (67,800) that had 75% or more of their enrollments reported as being delivered virtually. Examination of pass rates showed that students who had fewer than 25% of their enrollments delivered virtually and those who had 75% or more of their enrollments delivered virtually, outperformed the students in the middle two quartile groups. See Table D14. Table D15 and Table D16 show how the percentage of students, enrollments, and pass rates changed for LEA schools and PSA schools, respectively.

By Mobility Status

For the fourth consecutive year, mobility data were included as part of the data set. The mobility variable included the following statuses: stable, incoming, or outgoing. According to MI School Data, a student is marked as stable if he or she is in the same school for all collections for the school year, incoming students are those who transferred any time after the fall count day, and mobile students were present for fall count day but not subsequent ones. Some of the enrollments did not include information on this variable and were listed in the data tables as “Missing.” More information about this variable is available on the MI School Data Student Mobility page. Click on the About this Report down arrow on that page and then click About the Data to view definitions.

When it came to district stability, over 840,000 (83%) of the virtual enrollments were classified as stable. The pass rate for stable enrollments was 69%. Incoming enrollments to a district represented 7% of the virtual enrollments and had a pass rate of 49%. See Table D17.

When looking at mobility from a poverty perspective, we get a more nuanced picture. Eighty percent of virtual enrollments from students in poverty were stable compared to 90% for students who were not in poverty. The pass rate for stable, in poverty enrollments was 64% but rose to 80% for stable, not in poverty enrollments. For incoming virtual enrollments, there was a nine-percentage point advantage for students who were not in poverty (48% v. 57%). See Table D18.

Looking at mobility from a locale perspective showed somewhat similar virtual enrollment percentages across geographies. Rural schools had the lowest percentage of stable enrollments at 82%. Town schools were next at 83% followed by City schools at 84%. Suburban schools reported 85% of their enrollments as stable. See Table D19. Virtual pass rates showed a similar pattern. Stable enrollments from Rural schools had a 69% pass rate whereas the pass rate was 73% for Suburban schools. The incoming pass rates tended to lag the stable pass rates regardless of the locale. See Table D20.

A final mobility dimension explored was how enrollment and performance varied across full-time and part-time virtual schools. Full-time virtual or cyber schools had a lower percentage of their virtual enrollments designated as stable compared to part-time (77% v. 88%). The full-time pass rate for stable enrollments also lagged that of students from part-time virtual programs (66% v. 71%). See Table D21.

By Non-Virtual Course Performance

Part-time virtual learners with at least three non-virtual courses were classified into one of three categories based on their success in those non-virtual courses. The three categories were:

• Passed all Non-Virtual Courses
• Did Not Pass 1 or 2 Non-Virtual Courses
• Did Not Pass 3 or More Non-Virtual Courses

In total, 82% of part-time virtual learners had at least three or more non-virtual enrollments. Of that group, 45% of students passed all their non-virtual courses, 18% did not pass one or two, and 37% did not pass three or more. There were clear differences in virtual pass rates between the three categories. Students passing all their non-virtual courses had an 84% virtual pass rate. Students who did not pass one or two non-virtual courses had a virtual pass rate of 70%, and those with the lowest non-virtual success had a virtual pass rate of only 45%. See Table D22.

By Virtual Course Performance

Fifty-two percent of virtual learners passed every virtual enrollment they took. This was the same as the prior year. Seventeen percent did not pass any of their virtual enrollments, and 31% passed some, but not all of their virtual enrollments. Students who passed all their virtual courses were responsible for 37% of the virtual enrollments. Students with mixed success generated 48% of the enrollments, and students who did not pass any of their virtual courses accounted for 15% of the virtual enrollments (compared to 14% in 2022-23). See Table D23.

For the students who did not pass any of their virtual courses, 33% only took one or two virtual courses. On the other hand, over 15,000 students did not pass five or more virtual courses, and close to 5,000 students did not pass 11 or more virtual courses. See Table D24 and Table G11. This is the first time since the 2018-19 school year where the percentage of virtual learners who didn’t pass any virtual courses and failed 11 or more exceeded 3% of the population. Further analysis of students failing all their 11 or more virtual courses showed 89% of these students had a single school report data for them. Close to 70% of these students came from full-time virtual programs. Over 800 students used special education services (17%), and over 4,000 of these students (85%) were in poverty.

What Table G11 makes clear is that for students who do not pass any of their virtual enrollments, “withdrawns” were rampant. For the virtual enrollments from students who did not pass any of their virtual enrollments, 47% had a “Withdrawn” status (exited, failing, or passing), and another 24% were classified as “Incomplete.” For those taking 11 or more virtual courses, 39% had a “Withdrawn” status, and 36% were marked “Incomplete.” In each case, only 28% and 24% of the virtual enrollments, respectively, were actually classified as “Completed/Failed.” Please see the section on Pass Rate Calculations for more elaboration on the impact of such issues on pass rates.

By Virtual Usage

Continuing pre-pandemic trends, virtual learners had the highest pass rates when they took one or two virtual courses. Students taking one to two virtual courses had a pass rate of 80% compared to a pass rate of 73% for those taking three to four virtual courses and a pass rate of 61% for students taking five or more virtual courses. About 34% of students took one or two virtual courses; however, 54% were found to have taken five or more virtual courses during the year. See Table D25.

A new table in last year’s report, Table D26, shows pass rate by virtual method and virtual usage. Blended Learning students had the highest overall pass rate (68%) and specifically, those taking one to two virtual courses were among the highest pass rates at 79%. Students enrolled in one to two Online Courses were also among the highest pass rates at 80%. For the Digital Learning and Online Course virtual usage methods, pass rates decreased as virtual course usage increased.

State Assessment

By Subject Area

State assessment data can be used to provide an independent measure of student performance. Based on SAT and M-STEP data from students in 11th grade, virtual learners showed lower percentages reaching proficiency on the Evidence-Based Reading and Writing (SAT), Mathematics (SAT), Science (M-STEP) and Social Studies (M-STEP) examinations than the statewide proficiency rates. Forty-one percent of the 11th grade virtual learners tested proficient in Evidence-Based Reading and Writing, and 17% were proficient in Mathematics. For Science, 30% tested proficient whereas Social Studies had 32% of the virtual learners reach proficiency. See Table E1. The pattern was similar for those taking the 8th grade assessments. See Table E2.

By Non-Virtual Performance

As expected, the percentage of 8th and 11th grade virtual learners testing proficient on these state tests varied considerably when accounting for their non-virtual performance. For instance, students taking a minimum of three non-virtual courses and passing all of them had proficiency rates that exceeded the statewide average for each assessment. Students who did not pass one or two of their non-virtual courses and those not passing three or more of their non-virtual courses had much lower rates of proficiency. See Table E3 and Table E4.

By Poverty Status

Students in poverty consistently recorded proficiency rates that were considerably lower than their peers who were not in poverty. As examples, 27% of virtual learners in poverty scored proficient on the 11th grade Evidence-Based Reading and Writing exam compared to 59% for those who were not in poverty. For Mathematics, only 9% of 8th grade virtual learners in poverty scored proficient compared to 37% for those not in poverty. See Table E5 and Table E6.

By Full- or Part-Time Type

Both 8th and 11th grade students taking virtual courses in a part-time capacity had higher rates of proficiency on the assessments compared to full-time virtual learners. For some assessments, the gap was sizable. For instance, the difference was 13 percentage points for 11th grade Mathematics and 14 points for 8th grade Mathematics. See Table E7 and Table E8.

Maps

Berrien, Alpena-Montmorency-Alcona, and Muskegon Area ISDs/RESAs had over 20% of students in their service areas take a virtual course in 2023-24. In total, there were 10 ISDs/RESA with 15% or more of the students taking virtual courses. An additional 17 ISDs/RESA had at least 10% and less than 15% of their students take a virtual course. Only three ISDs/RESAs (Macomb, Livingston, and Manistee) had less than 5% of their students take at least one virtual course. See Figure 2.

Figure 2. 2023-24 Percentage of Students Who Took a Virtual Course (Non-Cyber) by ISD

Over one in four students (almost 5,700 students) attending a PSA cyber school resided within the Wayne RESA service area. The Genesee, Ingham, Kent, Macomb, and Oakland ISD service areas were the only other ISDs with 1,000 or more of their resident students attending PSA cyber schools. Forty-five of the 56 ISDs had 100 or more students attending a PSA cyber school. See Figure 3.

Figure 3. 2023-24 Count of PSA Cyber School Students by Resident ISD

Reflections on Higher Performing Schools

As the above sections of the report make clear, virtual learning performance, in general, has continued to produce mixed results. The analyses in this section will focus exclusively on those schools that achieved pass rates of 80% or higher to glean a clearer picture of what virtual learning looked like for these schools and programs and how it might have differed, if at all, from the state statistics.

There were 648 Michigan schools with virtual pass rates of 80% or higher, reflecting 46% of all schools in the state with virtual learners. These schools reported 56,860 virtual learners or about 37% of the state’s virtual learners. When zooming in on these higher performing schools, the data show:

• Successful virtual programs can support various numbers of students, enrollments, and courses offerings – These schools showed success with 10 or fewer students (35%) and 100 or more students (27%). See Table F1. Some offered few enrollments (116 schools had one to nine virtual enrollments) while others offered many (292 schools had 100 or more). See Table F2. They also varied in the number of course titles offered. Thirty-eight percent of these schools offered 10 or fewer virtual courses titles. Twenty-two percent had enrollments between 26 and 50 courses, and 17% of these schools had students in more than 50 different virtual courses. See Table F3.
• LEA and PSA schools can offer successful virtual programs – Forty-six percent of LEA schools with virtual programs had schoolwide virtual pass rates of 80% or higher. For PSA schools, 38% achieved pass rates of 80% or higher. See Table F4. Both traditional school districts and charter districts can run successful virtual programs.
• Schools in cities, suburbs, towns, and rural settings are proving virtual learning success – All locales had schools with virtual pass rates of 80% or higher. Rural and Suburban schools had almost half (47%) of their schools reach this threshold, and City schools were close behind at 45%. See Table F5. These schools are provide evidence of virtual learning success across the various geographies of the state.
• These schools show strong results across students of different race/ethnicities – These higher performing schools also showed promise for equitable outcomes for students of different races and ethnicities. The pass rates for African American or Black students (88%) and Hispanic or Latino (89%) were considerably closer to the White pass rate (91%) than it was across all schools. Asian students had the highest pass rate at 93%. See Table F6. For these schools, virtual programs appear to be approaching more equitable outcomes.
• Students in poverty are succeeding in these virtual programs – Recall that across the entire state, students in poverty had a pass rate (58%) that was 19 percentage points lower than those virtual students who were not in poverty. In these 648 schools, the virtual pass rate for students in poverty rose to 88%—considerably closer to the 93% virtual pass rate for the students in those schools who were not in poverty. In these higher performing schools, students in poverty continued to represent a large percentage of virtual learners (49%) and virtual enrollments (53%), but those percentages were considerably smaller than the 64% of virtual learners and 71% of virtual enrollments seen across all virtual programs across the state. See Table F7. Additionally, virtual program success varied by a school’s free or reduced-lunch category (FRL). Sixty-nine percent of Low FRL schools with virtual learners achieved virtual pass rates of 80% or higher. The virtual pass rate was 60% of the Mid-Low FRL schools, 40% of Mid-High FRL schools, and 30% of High FRL schools. See Table F8. While some High FRL schools showed it was possible, it was considerably rarer than it was for Low FRL schools.
• Both full- and part-time programs can run effective virtual programs, but success is rarer for full-time programs – Forty-seven percent of part-time programs were able to yield schoolwide virtual pass rates of 80% or higher. It was considerably more difficult for full-time programs to achieve similar success. Only 15 of the 76 full-time programs (20%) reached the 80% pass rate mark. See Table F9.
• Both general education and alternative education programs reached 80% school-wide virtual pass rates – There were 587 general education schools in Michigan that achieved schoolwide virtual pass rates of 80% or higher. These schools represented 52% of general education schools with virtual programs. For alternative programs, 50 schools reached this mark. As a percentage of alternative programs, it represented just 19% of such schools, indicating that while it is possible, this threshold of success remains a sizable challenge. See Table F10.
• Virtual students can perform at or above their face-to-face performance level – In these 648 schools, there were 11,664 virtual learners who took a minimum of three virtual courses and had data for a minimum of three non-virtual courses. Eighty percent of these students had virtual pass rates that met or exceeded their non-virtual pass rates. See Table F11.

Conclusion

This year’s report represents the 14th year of data on the effectiveness of virtual learning in Michigan’s K-12 system. Many trends witnessed in past years continue to exist.

Table 1. Summary of Virtual Learning Metrics by School Year Since 2010-11

As Table 1 makes clear, the huge influx of virtual learners during the pandemic has mostly subsided and levels seem to be more in line with pre-pandemic trends. Unfortunately, the reduction in virtual learners and enrollments has been accompanied by an 11-percentage point drop in the virtual pass rate since 2020-21.

As we predicted in last year’s report, the virtual pass rate continued to shift toward pre-pandemic levels. We have been monitoring several factors likely to impact the rate. The first is changes in schools entering and existing. One hundred forty-four new schools were represented in this year’s data while 198 schools from last year dropped out because they didn’t have any virtual learners this year. The new schools added close to 28,000 virtual enrollments, roughly the same amount as the departing schools contributed last year. Interestingly, the new schools outperformed the departing schools; new schools had a 77% pass rate this year while the schools that left had a 70% pass rate the prior year. Obviously, the weight of this change on the pass rate was minimal compared to schools who were in both last year and this year. There were 1,277 schools in both years. The pass rate for these schools dropped from 65% to 63%.

A second shift related to Alternative Education programs. Prior to the pandemic, Alternative Education programs produced close to half the virtual enrollments. At the height of the pandemic, they dropped to just 10%. Since then, the percentage has been rebounding; this year, alternative education enrollments rose back up to be 40% of the virtual enrollments, 4 percentage points higher than the prior year. This is particularly important because the pass rate gap between Alternative Education programs and General Education programs was sizable. For this year, that performance gap was 21 percentage points lower for Alternative Education programs.

A third dynamic to understand relates to the grade levels of virtual learners. Pre-pandemic, we saw about 80% of the virtual enrollments come from the high school level. That percentage dropped to 40% for the 2020-21 school year. This year, the high school percentage was up to 71% of virtual enrollments, three percentage points higher than the prior year. With the K-5 pass rates at 80% and the 6-8th grade pass rate at 70%, enrollments at these levels tend to prop up the overall virtual pass rate. For instance, if the elementary, middle school, and high school virtual pass rates remained the same, but the proportion of virtual enrollments were the same as the 2019-20 school year, we would have seen the virtual pass rate drop by about 1.5%. Therefore, if virtual enrollments continue to shift toward a larger percentage of high school enrollments, a slight decline in the overall pass rate is predictable.

Given the 2023-24 figures for these three key factors, we predict there is likely more correction coming. Thus, we believe the overall pass rate will continue to backslide. That said, there is optimism that the floor may be closer to 60% rather than the 55% that was consistently observed prior to the pandemic.

On the positive side, the report also captured examples of schools and students benefiting from virtual learning. Forty-six percent of virtual learners attended schools that had virtual pass rates of 80% or higher, and equity of outcomes was much closer to desired reality. Clearly, these schools add to the evidence that online learning can and does work for many schools and students. To date, however, these schools reflect more of the exception—the hope—rather than the rule. As school, community, and legislative leaders evaluate their virtual learning programs, the data provided in this report can serve as informative benchmarks, and the varied analyses can be used as models to understand local implementation success at a deeper level.

School leaders looking to take the next step forward with their virtual programs may find value in the many free resources that Michigan Virtual has authored. These resources include a series of practical guides to online learning designed for students, parents, teachers, mentors, school administrators, and school board members. Michigan Virtual also provides quality reviews of supplemental online learning programs to Michigan schools at no cost. There are also the National Standards for Quality Online Learning, which offer frameworks to evaluate online programs, online teaching, and online courses. Finally, educational leaders looking to communicate and collaborate with others around the future of learning may find value in the Future of Learning Council.

Appendix A – Methodology

COVID-19 Impact

Readers should note that the COVID-19 pandemic appears to have significantly impacted data from 2019-20 through the 2022-23 reports. It may still be impacting this year’s data; caution is advised when comparing this year’s findings with prior years.

About the Data

The data for this report came from the following sources:

• Michigan Student Data System – School Year 2023-2024;
• Educational Entity Master (EEM);
• Michigan Student Data System Teacher Student Data Link (TSDL) – Collection Year 2023-2024; and
• Michigan’s K-12 Virtual Learning Effectiveness Report, 2022-23 – Used for comparing this year’s data with the 2022-23 school year.

Because the data for this report incorporates a variety of sources, the findings within may differ from those found through the MI School Data portal which may use different query parameters.

Enrollments classified as virtual in this report were treated as such due to the TSDL virtual method field indicating virtual delivery. Enrollments where the TSDL virtual method field was set to “Blended Learning,” “Digital Learning,” or “Online Course” were treated as virtual. According to the Michigan Student Data System Collection Details Manual Version 4.0, the virtual method field indicates “the type of virtual instruction the student is receiving.” (See page 346).

In prior years of the report, additional strategies, such as keyword searches of the local course title field, were used to flag virtual enrollments. Past years demonstrated that such efforts yield a low percentage of the virtual enrollments. This effort was discontinued starting with the 2020-21 report.

Michigan Virtual Students

Because this report is published by Michigan Virtual, some people have falsely concluded that the data in this report is about Michigan Virtual students only. Quite the contrary, the data in this report represent K-12 virtual learning across all providers, and Michigan Virtual as a provider would reflect only a small percentage of the virtual enrollments covered in this report. Readers interested in Michigan Virtual specific results can find those published in its Annual Report: 2023-24, which include data on the number of students, districts, and enrollments served as well as its virtual pass rate.

Enrollment Calculations

Enrollment data for this report principally relies on data collected in the MSDS Student Course Component. See page 324 of the Michigan Student Data System Collection Details Manual Version 4.0 for more details about this collection. Through this collection, the State collects data for each course a student takes. It is important to note some key variations in the data collection that impact possible approaches to calculating enrollment counts.

An example of known variation is the local naming conventions for course titles. For instance, one school may call a course “English 9”, another “9th Grade English,” and yet another “ELA 9.” The Student Course Component resolves this issue by requiring schools to report each enrollment with a Subject Area Code and a Course Identifier Code (SCED Course Code). These codes are created by the National Center for Education Statistics through the School Courses for the Exchange of Data (SCED) initiative. By using these standardized codes, we can compare data more readily across schools.

Another important variation involves course sections. In addition to the course title and SCED Course Code, schools frequently parse a course title into multiple sections. For example, a school with trimester courses may break a course into three sections, one for each trimester. A semester-based school, on the other hand, may break up a course into two sections. Others have chosen to break their courses into even smaller units such as quarters while others report what seem to be course units or lessons. Sometimes, schools use course sections to differentiate the online and face-to-face components of courses. For our purposes, the key point is that there is not always one enrollment record per student per course title.

Multiple course sections for a single course title are not, in and of themselves, problematic. They could be resolved if a weighting variable—for instance, the fraction of a Carnegie unit each section represents—was collected. The State does collect a field, Credits Granted, in the Student Course Component that might be used. However, two main drawbacks significantly impair its use. The first is that the field is only required for Migrant-eligible and dual-enrolled students. As such, many enrollments do not have a reported value. The second hindrance is inconsistent reporting of data that do exist. In some cases, schools report the Carnegie unit that was possible to be earned (same value no matter the completion status of the enrollment), although others treat the field value as variable depending on how well the student did (e.g., report a 0.5 for a student with a “Completed/Passed” completion status, but a 0.0 for a student who had a “Completed/Failed” completion status). These drawbacks make the Credits Granted field unusable as a weighting variable.

The challenge of variable course sections reported is multiplied when more than one school entity reports on the same pupil. The data appear to contain instances of two or more schools reporting on the same enrollments. Flavors of this appear to be a school partnering with an ISD to provide special education services and both reporting the same enrollments. Another example appears to occur when a student transfers from one district and then enrolls in the same courses at the new school. Table A1 and Table A2 highlight enrollment variation.

Table A1. 2023-24 Virtual Enrollment Counts and Pass Rates by Number of Virtual Enrollments Per Student/SCED Code Pair

Table A2. 2023-24 Percentage of Students by Total Student Enrollment Counts (Virtual and Non-Virtual) and Full- or Part-Time Schools

Given these data limitations, enrollment counts and related data figures in this report should be treated as estimates that, generally speaking, convey the trends observed for the school year.

Pass Rate Calculations

For this report, the pass rate was calculated based on data reported in the “Completion Status” field. For more information about the Completion Status field, including definitions for each status, see page 341 of the Michigan Student Data System Collection Details Manual Version 4.0. Column one of Table A3 displays the various statuses reported by schools for the virtual enrollments.

Table A3. 2022-23 Number and Percentage of Virtual Enrollments by Completion Status

Throughout this report, the pass rate simply represents the percentage of virtual enrollments with a status of “Completed/Passed.” Notice that the percentage of enrollments with a “Completed/Passed” status in Table A3 matches the statewide pass rate. This pass rate formula remains consistent with past reports.

Please keep in mind that calculating the pass rate in this manner will result in the lowest possible percentage. To illustrate why this is, consider the completion status of “Audited.” These virtual enrollments are not “failures” per se, but act as such in the formula since they are added to the formula’s denominator without impacting the numerator. Another example is enrollments with a completion status of “Incomplete.” About 9% of the virtual enrollments in this report were classified as “Incomplete.” As such, they are treated in the report’s pass rate formula as zero passes, even though some may eventually be awarded a passing status. Finally, it is unclear how to best treat enrollments with a “Withdrawn” status. For instance, 3% of the virtual enrollments this year were marked as “Withdrawn/Passing,” meaning that the student was passing the course at the time the student was withdrawn. Should these enrollments be counted as failures? What about students whose enrollments were marked as “Withdrawn/Exited” (8% of the virtual enrollments)? Based on the data available, there is no way to determine whether that exiting occurred in the first few weeks of class or the final weeks of class. The data do not provide insight into whether the student was re-enrolled in a different course or whether it was too late for re-enrollment in a credit-bearing opportunity for the student.

The research team raises these issues because they represent questions for which there are no definitive answers. In the end, the team decided to report the pass rate as the percentage of all virtual enrollments that were reported as “Completed/Passed.” To provide readers with a better idea of the impact of this approach, additional data tables are provided in Appendix G that allow interested readers to draw their own conclusions and to calculate their own formulas for many of the pass rates reported.

Appendix B – School Tables

Note: Click on the hyperlinked table number to return to the section of the report that discusses the table.

Table B1. Two Year Comparison (2022-23 and 2023-24) of Virtual Enrollment Data with Pass Rates

Table B2. Virtual Enrollment Differences for Schools Reporting Virtual Learners in Both 2022-23 and 2023-24

Table B3. Virtual Pass Rate Differences for Schools Reporting Virtual Learners in Both 2022-23 and 2023-24

Table B4. 2023-24 K-12 Virtual Enrollment Data by Grade Level

Table B5. 2023-24 Pass Rate Comparison for Students in Their Virtual and Non-Virtual Courses by Grade Level

Table B6. 2023-24 Number and Percentage of Schools and Virtual Enrollments by School Pass Rate

Table B7. 2023-24 Number and Percentage of Schools and Virtual Enrollments by Entity Type

Table B8. 2023-24 Number of Virtual Enrollments and Virtual Pass Rate by Entity Type

Table B9. 2023-24 Number and Percentage of Full-Time (FT) Virtual or Cyber School by Entity Type

Table B10. 2023-24 Number and Percentage of Students and Enrollments from Full-Time (FT) Virtual or Cyber Schools with Virtual Pass Rates by Entity Type

Table B11. 2023-24 Number and Percentage of Part-Time (PT) Virtual Schools by Entity Type

Table B12. 2023-24 Number and Percentage of Students and Enrollments from Part-Time (PT) Virtual Schools with Virtual Pass Rates by Entity Type

Table B13. 2023-24 Number and Percentage of Schools and Virtual Enrollments by School Emphasis

Table B14. 2023-24 Number of Virtual Enrollments and Virtual Pass Rate by School Emphasis

Table B15. 2023-24 Virtual Pass Rates for General Education and Alternative Education Schools by Entity Type

Table B16. 2023-24 Number and Percentage of Schools and Virtual Enrollments by Number of Virtual Enrollments per School

Table B17. 2023-24 Percentage of Schools by Ratio of Virtual Courses to Student and School Pass Rate

Table B18. 2023-24 Number and Percentage of Schools and Virtual Enrollments by Locale

Table B19. 2023-24 Percentage of Schools with Virtual Enrollments by Virtual Enrollment Totals and Locale

Table B20. 2023-24 Virtual Pass Rate by Locale

Table B21. 2023-24 Percentage of Schools with Virtual Enrollments by Building Pass Rate and Locale

Table B22. 2023-24 Number and Percentage of Schools with Virtual Enrollments by School Free or Reduced-Price Lunch Categories

Table B23. 2023-24 Number and Percentage of Virtual Enrollments with Virtual Pass Rate by School Free or Reduced-Price Lunch Categories

Table B24. 2023-24 Number of Schools, Students, and Enrolls for Schools with 1,000 or More Virtual Enrollments with Courses Per Student and Pass Rate by Entity Type

Table B25. 2023-24 Number of Schools, Students, and Enrolls for Schools with Less Than 1,000 Virtual Enrollments with Courses Per Student and Pass Rate by Entity Type

Table B26. 2023-24 Number of Schools, Students, and Enrolls for Schools with 1,000 or More Virtual Enrollments with Courses Per Student and Pass Rate by School Emphasis

Table B27. 2023-24 Number of Schools, Students, and Enrolls for Schools with Less Than 1,000 Virtual Enrollments with Courses Per Student and Pass Rate by School Emphasis

Appendix C – Course Tables

Note: Click on the hyperlinked table number to return to the section of the report that discusses the table.

Table C1. 2023-24 Number of Schools, Virtual Students, and Virtual Enrollments with Percentage of Virtual Enrollments and Virtual Pass Rate by Subject Area

Table C2. 2023-24 Pass Rate Comparison for Students in Their Virtual and Non-Virtual Courses by Subject Area

Table C3. 2023-24 Number of Schools, Virtual Students, and Virtual Enrollments with Percentage of Virtual Enrollments and Virtual Pass Rate by Course Title for the Top 10 Most Enrolled in English Language and Literature Courses

Table C4. 2023-24 Number of Schools, Virtual Students, and Virtual Enrollments with Percentage of Virtual Enrollments and Virtual Pass Rate by Course Title for the Top 10 Most Enrolled in Mathematics Courses

Table C5. 2023-24 Number of Schools, Virtual Students, and Virtual Enrollments with Percentage of Virtual Enrollments and Virtual Pass Rate by Course Title for the Top 10 Most Enrolled in Life and Physical Sciences Courses

Table C6. 2023-24 Number of Schools, Virtual Students, and Virtual Enrollments with Percentage of Virtual Enrollments and Virtual Pass Rate by Course Title for the Top 10 Most Enrolled in Social Sciences and History Courses

Table C7. 2023-24 Number of Schools, Virtual Students, and Virtual Enrollments with Percentage of Virtual Enrollments and Virtual Pass Rate for AP Courses

Table C8. 2023-24 Virtual Enrollments Percentage by Locale and Subject Area

Table C9. 2023-24 Virtual Enrollment Pass Rates by Locale and Subject Area

Table C10. 2023-24 Number and Percentage of Virtual Enrollments with Pass Rates by Student Sex and Subject Area

Table C11. 2023-24 Number of Schools, Virtual Students, and Virtual Enrollments with Percentage of Virtual Enrollments and Pass Rate by Virtual Method

Appendix D – Student Tables

Note: Click on the hyperlinked table number ro return to the section of the report that discusses the table.

Table D1. 2023-24 Number of Schools and Virtual Students with Percentage of Virtual Students and Percent Year over Year Change

Table D2. 2023-24 Number of Schools, Virtual Students, Virtual Enrollments with Percentage of Virtual Students and Virtual Enrollments and Virtual Pass Rate by Student Sex

Table D3. 2023-24 Number of Schools, Virtual Students, Virtual Enrollments with Percentage of Virtual Students and Virtual Enrollments and Virtual Pass Rate by Race/Ethnicity

Table D4. 2023-24 Number of Schools, Virtual Students, Virtual Enrollments with Percentage of Virtual Students and Virtual Enrollments and Virtual Pass Rate by Poverty Status

Table D5. 2023-24 Pass Rate Comparison for Students in Their Virtual and Non-Virtual Courses by Poverty Status

Table D6. 2023-24 Percentage of Virtual Students and Virtual Enrollments in Poverty with Pass Rate by Virtual Type

Table D7. 2023-24 Number and Percentage of Virtual Students by Free or Reduced-Price Lunch Category

Table D8. 2023-24 Number of Schools, Virtual Students, Virtual Enrollments with Percentage of Virtual Students and Virtual Enrollments and Virtual Pass Rate by Special Education Status

Table D9. 2023-24 Number of Schools, Virtual Students, Virtual Enrollments with Percentage of Virtual Students and Virtual Enrollments and Virtual Pass Rate by Primary Disability

Table D10. 2023-24 Number and Percentage of Virtual Students Compared to All MI Students with IEPs by Primary Disability

Table D11. 2023-24 Number of Schools, Virtual Students, Virtual Enrollments with Percentage of Virtual Students and Virtual Enrollments and Virtual Pass Rate by Home-School/Nonpublic Student Status

Table D12. 2023-24 Number of Schools, Virtual Students, Virtual Enrollments with Percentage of Virtual Students and Virtual Enrollments and Virtual Pass Rate by Full- or Part-Time Status

Table D13. 2023-24 Pass Rate Comparison for Full- and Part-Time Virtual Students by Virtual Subset

Table D14. 2023-24 Number and Percentage of Virtual Students and Virtual Enrollments with Pass Rates by Students’ Percentage of Enrollments Taken Virtually

Table D15. 2023-24 Number and Percentage of Virtual Students and Virtual Enrollments from LEA Schools Only with Pass Rates by Students’ Percentage of Enrollments Taken Virtually

Table D16. 2023-24 Number and Percentage of Virtual Students and Virtual Enrollments from PSA Schools Only with Pass Rates by Students’ Percentage of Enrollments Taken Virtually

Table D17. 2023-24 Number of Schools, Virtual Students, Virtual Enrollments with Percentage of Virtual Students and Virtual Enrollments and Virtual Pass Rate by District Mobility

Table D18. 2023-24 Number and Percentage of Virtual Enrollments with Pass Rates by Known Poverty Status and District Mobility

Table D19. 2023-24 Percentage of Virtual Enrollments by Locale and District Mobility

Table D20. 2023-24 Virtual Pass Rates by Locale and District Mobility

Table D21. 2023-24 Percentage of Virtual Enrollments with Pass Rates by Full-Time (FT) or Part-Time (PT) Virtual Status and District Mobility

Table D22. 2023-24 Number and Percentage of Part-Time Virtual Students and Virtual Enrollments with Pass Rate by Non-Virtual Performance (Minimum of 3 Non-Virtual Enrollments)

Table D23. 2023-24 Number and Percentage of Virtual Students and Virtual Enrollments with Virtual Pass Rate by Virtual Course Performance

Table D24. 2023-24 Number and Percentage of Virtual Students Who Did Not Pass Any Virtual Courses by the Number of Virtual Courses They Took

Table D25. 2023-24 Number and Percentage of Virtual Students and Virtual Enrollments with Pass Rates by Virtual Usage

Table D26. 2023-24 Virtual Pass Rate by Virtual Method and Virtual Usage

Appendix E – State Assessment Tables

Note: Click on the hyperlinked table number to return to the section of the report that discusses the table.

Table E1. 2023-24 Comparison of Virtual and State Proficiency Rates on 11th Grade State Assessment Measures

Table E2. 2023-24 Comparison of Virtual and State Proficiency Rates on 8th Grade State Assessment Measures

Table E3. 2023-24 11th Grade State Assessment Proficiency Rates for Virtual Learners with Three or More Non-Virtual Enrollments by Non-Virtual Performance

Table E4. 2023-24 8th Grade State Assessment Proficiency Rates for Virtual Learners with Three or More Non-Virtual Enrollments by Non-Virtual Performance

Table E5. 2023-24 11th Grade State Assessment Proficiency Rates for Virtual Learners by Poverty Status

Table E6. 2023-24 8th Grade State Assessment Proficiency Rates for Virtual Learners by Poverty Status

Table E7. 2023-24 11th Grade State Assessment Proficiency Rates for Virtual Learners by Virtual Type

Table E8. 2023-24 8th Grade State Assessment Proficiency Rates for Virtual Learners by Virtual Type

Appendix F – Higher Performing Schools Tables

Note: Click on the hyperlinked table number to return to the section of the report that discusses the table.

Table F1. 2023-24 Number and Percentage of Schools with 80% or Higher Pass Rate by Virtual Learner Count Category

Table F2. 2023-24 Number and Percentage of Schools and Virtual Enrollments from Schools with 80% or Higher Pass Rate by Virtual Count Category

Table F3. 2023-24 Number and Percentage of Schools with 80% or Higher Pass Rate by Number of Virtual Courses Offered

Table F4. 2023-24 Number and Percentage of Schools with 80% or Higher Pass Rate as a Percentage of All Virtual Schools

Table F5. 2022-23 Number and Percentage of Schools with 80% or Higher Pass Rate by Locale

Table F6. 2022-23 Number of Students and Virtual Enrollments with Pass Rate Data from Schools with 80% or Higher Pass Rate by Race/Ethnicity

Table F7. 2023-24 Number of Students and Virtual Enrollments with Pass Rate Data from Schools with 80% or Higher Pass Rate by Poverty Status

Table F8. 2023-24 Number and Percentage of Schools with 80% or Higher Pass Rate by School Free or Reduced-Price Lunch Category

Table F9. 2023-24 Number and Percentage of Schools with 80% or Higher Pass Rate by Full- or Part-Time Status

Table F10. 2023-24 Number and Percentage of Schools with 80% or Higher Pass Rate by School Emphasis

Table F11. 2023-24 Number and Percentage of Students* from Schools with 80% or Higher Pass Rate by Pass Rate Difference Category

Appendix G – Completion Status Tables

Note: Click on the hyperlinked table number to return to the section of the report that discusses the table.

Table G1. 2023-24 Number and Percentage of Virtual Enrollments by Completion Status

Table G2. 2023-24 Percentage of Virtual Enrollments by Completion Status and Entity Type

Table G3. 2023-24 Number and Percentage of Full-Time Virtual Enrollments by Completion Status

Table G4. 2023-24 Number and Percentage of Part-Time Virtual Enrollments by Completion Status

Table G5. 2023-24 Percentage of Virtual Enrollments by Completion Status and School Emphasis

Table G6. 2023-24 Percentage of Virtual Enrollments by Completion Status and Core Subject Area

Table G7. 2023-24 Percentage of Virtual Enrollments by Completion Status and Student Sex

Table G8. 2023-24 Percentage of Virtual Enrollments by Completion Status and Race / Ethnicity

Table G9. 2023-24 Number and Percentage of Virtual Enrollments by Completion Status and Poverty Status

Table G10. 2023-24 Number and Percentage of Virtual Enrollments by Completion Status and Special Education Status

Table G11. 2023-24 Percentage of Virtual Enrollments by Completion Status for Students Who Did Not Pass Any of Their Virtual Courses

Introduction

The integration of technology in education has been a subject of debate and research for over a century, with educators’ resistance to adoption persisting despite rapid technological advancements. This historical context is crucial for understanding the current landscape of technology adoption in education, particularly as we face the emergence of artificial intelligence (AI) in educational settings. Much like the resistance faced by calculators and computers in the past, AI is encountering similar skepticism and barriers to adoption. This literature review aims to conduct an extension of a meta-analysis of recent studies to identify key predictors of technology adoption among educators, with the goal of informing targeted training programs that address the strongest predictors of acceptance.

The foundations of modern educational psychology, established by pioneers like Jean Piaget in the early 20th century, emphasized the importance of active, constructive learning experiences (Piaget, 1936). This perspective naturally aligns with the potential of technological tools, including AI, to provide rich, interactive learning environments. Later, works such as “How People Learn” (Bransford et al., 2000) further highlighted technology’s potential to support effective learning principles, principles that are now being extended to AI-enhanced educational tools.

Despite these theoretical underpinnings, the history of educational technology adoption reveals a pattern of resistance. Cuban (1986) documented cycles of enthusiasm and disappointment surrounding various technologies introduced in classrooms throughout the 20th century. This resistance has persisted into the digital age, with Ertmer (1999) identifying both external and internal barriers to technology integration in classrooms. Concerns about efficacy, job displacement, changing teacher roles, and the constant pressure to keep up with technological advancements have all contributed to ongoing resistance (Selwyn, 2011; Howard, 2013; Zhao & Frank, 2003). In addition, resistance is further complicated by lack of pedagogical training and administrator support, which push teachers to keep the status quo rather than ride the wave of each “new and next” educational gimmick (Michigan Virtual, 2024).AI, as the latest iteration of educational technology, is facing similar challenges.

The rapid evolution of technology in recent decades, culminating in the advent of AI, has intensified the challenges of adoption. The rise of artificial intelligence in education has sparked a new wave of apprehension among teachers, reminiscent of past technological innovations, and it’s not unwarranted; recent studies, including work from Michigan Virtual (2024), have shown that educators express significant concerns about AI’s role in the classroom, specifically around inappropriate use of AI, ethical concerns, and student overreliance on AI. Additionally, another Michigan Virtual study by McGehee (2024) indicates that students simply using AI doesn’t make much of a difference for students, but rather how they use it is what makes an impact, indicating that thoughtful and appropriate integration is key to success.

A Walton Family Foundation (2024) study also found that many teachers express less than supportive views of AI and that distrust and unfamiliarity increase with age. Rand’s (2024) study of K-12 educators found that many teachers still aren’t using AI despite the large adoption rates in other industries, though many districts planned to train teachers to do so by the end of the 2023-24 school year. Trust et al. (2023) found that teachers worry about AI’s potential to replace human instruction, erode critical thinking skills, and exacerbate academic dishonesty.

These concerns echo the historical pattern of resistance to new educational technologies documented by Cuban (1986).

Moreover, the rapid development and deployment of AI tools have left many educators feeling unprepared and overwhelmed. Zhai et al. (2021) reported that teachers often feel they lack the necessary skills and knowledge to effectively integrate AI into their teaching practices. This technological anxiety is compounded by ethical concerns surrounding AI, such as data privacy and the potential for bias in AI systems (Holmes et al., 2022). The perceived complexity of technology, identified as a strong predictor of resistance in previous studies (Mac Callum et al., 2014), is particularly pronounced in the case of AI adoption (McGehee, 2023).

Despite these concerns, there is also a growing recognition of AI’s potential to enhance education. Zawacki-Richter et al. (2019) highlighted AI’s capacity to personalize learning experiences and provide valuable insights into student performance, and recent work by Michigan Virtual (McGehee, 2024) has shown that specific usage types of AI are associated with significantly higher student achievement.

However, the realization of these benefits hinges on addressing teachers’ apprehensions and providing adequate support and training (McGehee, 2023; Michigan Virtual, 2024). As Selwyn (2019) argues, the successful integration of AI in education will require a careful balance between technological innovation and pedagogical wisdom, with teachers playing a central role in shaping AI’s implementation in the classroom.

Understanding this historical context is crucial for setting the stage for the literature analysis of recent research on predictors of technology adoption among educators. By examining studies from the past two decades, the aim is to identify the key factors that influence educators’ willingness and ability to integrate technology, including AI, into their teaching practices. These predictors may include, but are not limited to, teacher attitudes, institutional support, professional development opportunities, and perceived usefulness of technology.

This literature meta-analysis will synthesize findings from multiple studies to provide a comprehensive view of the current state of technology adoption in education. By understanding these predictors, educational leaders and policymakers can develop more effective strategies to support and encourage technology integration, including AI, addressing the persistent challenges that have characterized this field for over a century.

This literature meta-analysis will proceed as follows: First discussed is the outline of the methodology for selecting and analyzing relevant studies. Next, is the presentation of our findings on the most significant predictors of technology adoption among educators. Finally, the implications of these findings for educational practice and policy will be discussed, considering how they relate to the historical context of resistance to technology adoption in education, with a particular focus on preparing educators for the integration of AI in their teaching practices.

Method

This literature meta-analysis uses the following predictors of large categorical factors of technology adoption:

• Perceived Ease of Use and Perceived Usefulness 
• Behavioral Intention
• Moderating Factors 

These predictor categories are taken from Scherer, Siddiq, & Tondeur’s (2019) meta-analysis and Granic’s (2022) meta-analysis on technology acceptance studies that largely used the Technology Acceptance Model or TAM instrument(s) (Davis, 1989) to study educational technology adoption. While not all studies included in this analysis used the TAM or its iterations, they observed or studied factors that are included in it or similar to it and were thus considered appropriate for inclusion. Factors not included in the TAM, and were not similar enough to be included as a TAM factor, are discussed separately.

This compilation is intended to be an extension of Scherer, Siddiq, & Tondeur’s, and Granic’s analyses, by including further studies and resources focused on AI in an effort to provide insight on how to properly support educators transitioning into the age of Artificial Intelligence.

Criteria for Inclusion and Analytical Methods

Each study that was selected was published within the last 20 years, uses teachers as the primary population, and focuses on technology adoption. All of the studies in Granic’s 2022 meta-analysis and Scherer, Siddiq, & Tondeur’s (2019) meta-analysis were included, with an addition of 16 AI technology adoption studies and 30 other general Edtech adoption studies.

Excluding the studies pulled from Grannic (2022) and Scherer, Siddiq, & Tondeur (2019), AI tools were used to find and organize resources that were relevant to this meta-analysis in addition to Google Scholar and the Google search engine. 

Search terms included: Artificial intelligence and Technology Acceptance Model, AI and TAM, AI, educators, and TAM, AI and educator acceptance, AI tools in education, adopting AI tools in education, AI tools and teacher adoption, teacher technology adoption, teachers and TAM, educators, and TAM.

The AI tools that were used for this meta-analysis extension were:

• Chat GPT (4.0) – used for determining the strength and impact of studies, as well as identifying potential studies for inclusion
  • This was the primary tool used for assistance in the study regarding analyses – all other tools were used as checks against ChatGPT for accuracy, clarity, and reasoning to determine if there were any severe discrepancies).
• Claude 3.5 (Anthropic) – used for determining the strength and impact of studies, as well as identifying potential studies for inclusion; additionally, it was used to assist in synthesizing findings for use in a summary table.
  • This was a supporting tool
• Google Gemini – used for determining the strength and impact of studies, as well as identifying potential studies for inclusion
  • This was a supporting tool
• Elicit – primarily used for identifying potential studies for inclusion; information from relevant studies was taken from here and imported into ChatGPT, Claude, and Gemini.
  • This was the primary AI tool used to identify potential studies, as well as extract information from studies. It is important to note that while this tool identifies specific information for studies and extracts it, the researcher still looked at each study and confirmed the extracted information.
• Scite.ai – primarily used for identifying potential studies for inclusion; information from relevant studies was taken from here and imported into ChatGPT, Claude, and Gemini.
  • This was the secondary AI tool used to identify potential studies and to extract information from studies. It is important to note that while this tool identifies specific information for studies and extracts it, the researcher still looked at each study and confirmed the extracted information.

While fewer in number, studies that specifically dealt with AI as the given technology adoption were given special attention and double weight, considering that AI tools are the newest form of educational technology to be rapidly dispersing and permeating schools for teachers and students alike. This weighting was only used when comparing Gen EdTech adoption vs AI adoption, because there are far more existing studies regarding general technology adoption as compared to AI adoption, thus it was easier to make comparisons between the two.

When analyzing the studies that were selected for this meta-analysis, three categories were calculated on key variables that each study dealt with strength, impact, and amount of evidence.

The criteria and definitions for the strength, impact, and amount of evidence categories are as follows:

• Strength – this is synonymous with the significance of (p-values) statistical hypothesis testing, correlation coefficients, and regression weights and coefficients of the variable across the studies it is concerned with. This is categorized as strong, moderate, or weak based on the findings across the number of supporting studies. These were taken verbatim, directly from the studies and then averaged by AI.
  • Strong – p values of less than .001, large F or t values, correlation coefficients of approximately [.6] or higher, large regression coefficients, and/or importance scores from models.
  • Moderate – p values of less than .05., moderate size F or t values, correlation coefficients of [.3] – [.5], moderate-sized regression coefficients, and/or importance scores from models.
  • Weak – p values very close to .05 or approaching statistical significance, small F or t values, correlation coefficients of less than [.3], and small regression coefficients and/or importance scores from models.
  • This was a categorical organization of studies by the researcher, followed by a comparison of frequencies of studies in each category by AI, resulting in an overall descriptor.
    • For example, if 10 studies reported strong relationships between two variables, and 4 reported moderate relationships, ChatGPT was asked to give an overall rating of the variable and would categorize it as strong because there are more than double the number of studies showing strong correlations or relationships than moderate, and none showing weak relationships.
• Impact- this is concerned with effect sizes, population size, and the strength of each predictor across each study in its group. These are categorized into Very Low, Low, Moderate, High, Very High, and Essential categories, sometimes with a direction of positive or negative if the findings across all of the studies provide enough consensus.
  • This was similar to Strength but had to do with population size and effect sizes of results in studies included in the analysis. Studies that dealt with differences between groups had their impact descriptors directly pulled from the text and were categorized into the 6 bins mentioned above. In addition, the larger the sample size, the more weight the studies were given by AI. Then an overall categorization was calculated by ChatGPT by comparing the amount of studies in each category and their weights.
  • As with the previous category of Strength, this was a categorical organization of studies by the researcher, followed by a comparison of frequencies of studies in each category by AI, resulting in an overall descriptor.
    • For example, if 5 studies with large effect sizes (each with large N), and a strength score of high, then 5 studies with moderate effect sizes (small N) and a strength score of high, and finally 5 studies with small effect sizes (small N) and a strength score of moderate were included, then ChatGPT would take this information and categorize this variable likely into the High category, based on the frequencies of the characteristics. 
• Amount of Evidence – this is concerned with how many studies support the finding. These are categorized into:
  • Low – fewer than 6 large N studies or fewer than 8 studies/resources with low N
  • Moderate – between 8 to 12 studies/resources, with a minimum of one high N resource or experimental/quasi-experimental methods
  • High – 12+ studies, or more than 8 studies/resources, with two or more high N or experimental/quasi-experimental methods
• AI – Evidence – this is concerned with only the number of studies that support the predictor regarding educator adoption of Artificial Intelligence only
  • This category is specific to evidence that deals with AI adoption as the technology in question. This is a numerical value of the number of studies that support the factor

It is important to note that these results were also compared to the findings from the previous two meta-analyses. It is intended to be an extension of, and thus, any discrepancies between this study and the other two will be discussed.

Findings

Using the methods described above and replicating the categories from the TAM and Granic’s (2022) work, a summary table presents the findings from this meta-analysis extension.

The table of findings (Table 1) and subsequent text presents a comprehensive overview of factors influencing technology acceptance among educators based on a meta-analysis of the subject. The factors are categorized into three main groups: antecedents of Perceived Ease of Use (PEU) and Perceived Usefulness (PU), antecedents of Behavioral Intention (BI), and moderating factors, each of which are defined as follows:

• Perceived Ease of Use
  • This refers to the degree to which a person believes that using a particular technology would be free of effort. It’s about how easy the user thinks it will be to learn and use the technology.
• Perceived Usefulness
  • This is defined as the degree to which a person believes that using a particular technology would enhance their job performance or life in general. It’s about how beneficial or valuable the user thinks the technology will be.
• Behavioral Intention
  • This refers to a person’s readiness to perform a given behavior. In the context of TAM, it’s the likelihood that a person will adopt and use the technology.
• Moderating Factors
  • This refers to variables that can influence the main effect of other factors. This means the strength of certain factors can vary based on these.
• The categories of Perceived Ease of Use (PEU) and Perceived Usefulness (PU) influence the user’s attitude toward using the technology, which in turn affects their Behavioral Intention (BI) to use it. The Behavioral Intention (BI) is what ultimately determines a person’s actual usage of a technology. 
• Therefore, when interpreting the antecedents or factors that influence PEU and PU, it is important to understand that they are essentially once removed from influencing the actual adoption and use of a technology because they are factors that have relationships with factors that influence adoption. Factors of BI are not considered once removed because BI essentially is technology adoption, and therefore any factor that influences this is closer in relationship than any PEU or PU Factor. 
• Moderating factors are just that; they are variables that have relationships with the strength of a factor depending upon the level of the moderating factor, an example of this would be Age moderates self-efficacy, which means that self-efficacy has significant differences depending on a participant’s age

Table 1 – Summary Table

General TAM Model Factors – Gen Edtech Adoption

Among the PEU (Perceived Ease of Use) and PU (Perceived Usefulness) antecedents, several factors stand out as particularly influential, consistent with the previous findings of Grannic (2022) and Scherer, Siddiq, & Tondeur’s (2019) work.

Self-efficacy, an individual’s judgment of their capability to use technology, shows a strong strength with a very large impact and high amount of evidence. This underscores the critical role of teachers’ confidence in their technological abilities (Scherer & Teo, 2019; Holden & Rada, 2011; Joo et al., 2018; Chuang et al., 2020; Leem & Sung, 2019; Celik & Yesilyurt, 2013; Wang et al., 2021).

Perceived Enjoyment and Technological Complexity both demonstrate strong strength with large impacts and high evidence, highlighting the importance of user experience in technology adoption. Perceived Enjoyment is shown to significantly enhance the likelihood of adoption when technology is engaging and enjoyable to use (Cheung & Vogel, 2013; Teo & Noyes, 2011; Padilla-Meléndez et al., 2013; Mun & Hwang, 2003; Chang et al., 2017; Leem & Sung, 2019). Technological Complexity, meanwhile, indicates that as complexity increases, adoption likelihood decreases unless sufficient support is provided (Celik & Yesilyurt, 2013; Aldunate & Nussbaum, 2013; Calisir et al., 2014; Hsu & Chang, 2013; Hanif et al., 2018; Tarhini et al., 2014).

Facilitating Conditions, which refer to resources and technology factors affecting usage, are deemed essential with strong strength and high evidence. This emphasizes the crucial role of institutional support and infrastructure in promoting technology adoption (Venkatesh et al., 2003; Fathema et al., 2015; Moran et al., 2010; Teo et al., 2016; Lawrence & Tar, 2018; Chiu, 2021; O’Bannon & Thomas, 2014).

Anxiety, on the other hand, shows a strong negative impact, indicating that teachers’ apprehensions about technology can significantly hinder adoption (Celik & Yesilyurt, 2013; Calisir et al., 2014; Howard, 2013; Chang et al., 2017; Chiu, 2021; Vongkulluksn et al., 2018).

In terms of Behavioral Intention antecedents, Self-efficacy again emerges as a critical factor with essential positive impact and high evidence. This reinforces the importance of building teachers’ confidence in their technological abilities (Scherer & Teo, 2019; Holden & Rada, 2011; Joo et al., 2018). Subjective Norm and Perceived Playfulness both show moderate strength but large impacts, suggesting that social influences and intrinsic motivation play significant roles in shaping intentions to use technology (Cheung & Vogel, 2013; Tarhini et al., 2014; Teo et al., 2018; Padilla-Meléndez et al., 2013; Scherer & Teo, 2019; Park et al., 2012; Vongkulluksn et al., 2018).

The moderating factors provide additional nuance to understanding technology acceptance. Age moderately affects the relationships between several key factors and behavioral intention (Tarhini et al., 2014; Wang et al., 2009; O’Bannon & Thomas, 2014; Scherer & Teo, 2019; Wong et al., 2012), while Gender shows a weaker influence (Tarhini et al., 2014; Wong et al., 2012; Teo & van Schaik, 2012; Park et al., 2012). Individual-level Cultural Values demonstrate moderate strength and impact, suggesting that cultural context plays a role in technology acceptance (Teo et al., 2008; Nistor et al., 2013; Sánchez-Prieto et al., 2020; Chuang et al., 2020). Notably, Technological Innovation strongly moderates the relationships between subjective norm, perceived usefulness, and behavioral intention, highlighting the importance of keeping pace with evolving technologies in education (Venkatesh et al., 2003; Moran et al., 2010; Leem & Sung, 2019; Teo et al., 2021).

Below are two figures that visualize the strengths (Figure 1) and the impacts (Figure 2) of each variable in the TAM according to the AI assisted analysis. It is important to note that the values used to make these comparisons were calculated by AI, as described in the methodology section, but are summarized into categorical bins in Table 1.

Figure 1 – General EdTech Adoption factor strengths

Figure 2 – General EdTech Adoption factor impacts

Non-TAM Factors – Gen EdTech Adoption

Perceived Risk, Expectancy, and Perceived Trust are important predictors of teachers’ technology adoption, closely related to but distinct from the traditional predictors and their antecedents in the Technology Acceptance Model (TAM). The TAM primarily focuses on two key broad categorical predictors: Perceived Usefulness and Perceived Ease of Use, which explain users’ acceptance of technology based on its utility and the effort required to use it; these larger constructs, again, can be predicted in turn by smaller groups of antecedents.

Perceived Risk involves the potential negative consequences or uncertainties that teachers associate with adopting new technology. Unlike the TAM, which doesn’t explicitly consider risk, Perceived Risk addresses teachers’ concerns about the reliability of technology, privacy issues, and the possibility of failure or negative outcomes (Howard, 2013; Teo et al., 2016). High perceived risks can deter educators from integrating technology into their teaching practices, as they might worry about disruptions, loss of classroom control, or being judged negatively by colleagues or administrators. By acknowledging these fears, schools can reduce perceived risks through reliable support and training, encouraging greater technology adoption.

Expectancy, similar to Perceived Usefulness in the TAM, refers to teachers’ beliefs about the likelihood that technology will lead to positive outcomes, such as enhanced instructional effectiveness or improved student learning. However, Expectancy expands beyond just usefulness by incorporating elements of motivation and anticipated success (Hanif et al., 2018; Chang et al., 2017; Moran et al., 2010; Chen et al., 2008; Davis, 1989; Teo & van Schaik, 2012). When teachers believe that technology can make their jobs easier or more engaging for students, they are more likely to use it. This aligns with the principles of the Expectancy Theory, which suggests that individuals are motivated to adopt behaviors they expect to lead to desired outcomes.

Perceived Trust is another crucial factor that goes beyond the traditional TAM components. It encompasses teachers’ confidence that the technology is reliable, secure, and capable of performing as needed, as well as trust in the organizations or institutions providing it(Teo et al., 2018; Scherer et al., 2021; Celik & Yesilyurt, 2013). In the TAM framework, trust is not explicitly addressed, yet it plays a significant role in technology adoption, particularly in environments like schools where ethical considerations and professional standards are paramount. When teachers trust the technology and its providers, they are more likely to adopt it, knowing it aligns with their educational goals and responsibilities. Building trust through consistent positive experiences and transparent communication about technology’s capabilities and limitations can significantly enhance adoption rates.

AI Adoption Factors

The adoption of AI tools by teachers in educational settings is influenced by a multifaceted array of factors that both align with and extend beyond the traditional components of the Technology Acceptance Model (TAM). While TAM emphasizes Perceived Ease of Use and Perceived Usefulness as central predictors of technology acceptance, the integration of AI-based educational technology requires a deeper examination of additional variables that reflect the unique characteristics and challenges associated with AI technologies (Chocarro et al., 2021; Wang et al., 2021).

Self-efficacy is a significant factor influencing AI tool adoption, showing strong evidence as an antecedent of both perceived ease of use and behavioral intention (Chatterjee & Bhattacharjee, 2020; Zhang et al., 2023; Ayanwale et al., 2022; Nja et al., 2023; Alhumaid et al., 2023). An individual’s confidence in their ability to use technology effectively is crucial; teachers who believe they possess the necessary skills to navigate AI tools are more likely to perceive these tools as user-friendly and are consequently more inclined to integrate them into their teaching practices (Choi, Jang, & Kim, 2022; Woodruff, Hutson, & Arnone, 2023; Nazaretsky, Cukurova, & Alexandron, 2021). This is consistent with the TAM, where self-efficacy directly contributes to perceived ease of use (Sánchez-Prieto et al., 2019; Zhang et al., 2021).

System Accessibility and Technological Complexity also emerge as critical determinants in the adoption of AI-based EdTech, closely mirroring the TAM’s focus on ease of use (Nazaretsky et al., 2021; Rico-Bautista et al., 2021; Chocarro, Cortiñas, & Marcos-Matás, 2021; Woodruff et al., 2023). The ease of access to AI tools and the simplicity with which they can be operated significantly influence their perceived usability (Wang et al., 2021; Nja et al., 2023; Al Darayseh, 2023). Technologies that are accessible and uncomplicated encourage adoption by minimizing the perceived effort required to use them, aligning with the TAM’s premise that simplicity enhances user acceptance (Sánchez-Prieto et al., 2019; Zhang et al., 2023; Cukurova et al., 2023).

Subjective Norms and Perceived Playfulness indicate moderate influence on both behavioral intentions and perceived usefulness (Wang et al., 2021; Chocarro et al., 2021; Al Darayseh, 2023; Ayanwale et al., 2022). Subjective norms, which refer to the influence of colleagues, administrators, and the broader educational community, can significantly shape teachers’ attitudes toward adopting AI tools (Choi et al., 2022; Zhang et al., 2021). If there is a positive perception of AI within these social circles, teachers are more likely to view these tools as beneficial, thus enhancing their perceived usefulness (Nazaretsky et al., 2021; Cukurova et al., 2023; Alhumaid et al., 2023). Perceived playfulness, or the intrinsic enjoyment derived from using AI tools, adds another layer to technology adoption, highlighting the importance of motivational factors that go beyond mere utility (Chocarro et al., 2021; Zhang et al., 2021; Nja et al., 2023). This suggests that in educational contexts, enjoyment and engagement can be critical for sustaining technology use, expanding the TAM’s traditional focus on functionality and ease (Woodruff et al., 2023; Chatterjee & Bhattacharjee, 2020).

Ethical Issues and Transparency are particularly relevant in the context of AI adoption and are not explicitly covered by TAM (Choi et al., 2022; Nazaretsky et al., 2021; Rico-Bautista et al., 2021). The integration of AI tools in education raises concerns about biases in AI algorithms, student data privacy, and the transparency of AI decision-making processes (Cukurova et al., 2023; Al Darayseh, 2023; Zhang et al., 2023). Teachers may hesitate to adopt AI tools if they perceive them as ethically questionable or if there is a lack of clarity about how decisions are made by these technologies (Nazaretsky et al., 2021; Choi et al., 2022; Alhumaid et al., 2023). This perceived lack of transparency can undermine trust, which is essential for the adoption of AI-based EdTech (Nja et al., 2023; Zhang et al., 2021). Addressing these concerns by providing clear, transparent explanations of AI functionality and ensuring ethical standards are upheld is crucial for fostering trust and acceptance (Rico-Bautista et al., 2021; Chatterjee & Bhattacharjee, 2020).

Anxiety about using AI technologies represents another factor that differs from traditional TAM components (Ayanwale et al., 2022; Chatterjee & Bhattacharjee, 2020; Woodruff et al., 2023). Anxiety reflects personal traits that cause apprehension or fear when engaging with new technology. This factor can have both direct and inverse effects on technology adoption (Al Darayseh, 2023; Cukurova et al., 2023; Nazaretsky et al., 2021). Teachers who experience high levels of anxiety may perceive AI tools as difficult to use or may doubt their usefulness, hindering adoption (Woodruff et al., 2023; Zhang et al., 2021; Choi et al., 2022). Conversely, with adequate support and training, anxiety can be mitigated, enhancing perceived ease of use and fostering a more positive attitude toward AI tools (Wang et al., 2021; Nja et al., 2023; Alhumaid et al., 2023).

Cost and Time considerations also play a significant role in the adoption of AI-based EdTech, adding another dimension not explicitly covered by TAM (Cukurova et al., 2023; Alhumaid et al., 2023; Rico-Bautista et al., 2021; Nazaretsky et al., 2021). Teachers are more likely to adopt AI tools if they perceive them as cost-effective and time-saving (Wang et al., 2021; Zhang et al., 2023; Nja et al., 2023; Al Darayseh, 2023). However, if the adoption of these tools is seen as requiring substantial financial investment or adding to teachers’ workloads without clear benefits, resistance is likely (Woodruff et al., 2023; Alhumaid et al., 2023; Choi et al., 2022). This factor includes both the initial time and cost to learn and implement AI tools, as well as ongoing maintenance and the need for continuous professional development (Cukurova et al., 2023; Chatterjee & Bhattacharjee, 2020).

The factor, Required Shift in Pedagogy, is a further consideration influencing AI adoption, extending beyond the traditional TAM framework (Nja et al., 2023; Al Darayseh, 2023; Choi et al., 2022; Woodruff et al., 2023). AI tools often necessitate a change in teaching methods, requiring educators to rethink how they deliver content and assess student learning (Cukurova et al., 2023; Chatterjee & Bhattacharjee, 2020; Zhang et al., 2023). The perceived need to modify pedagogical practices can create resistance, particularly among teachers accustomed to traditional methods, even if the perceived usefulness and ease of use of AI tools are high (Nazaretsky et al., 2021; Alhumaid et al., 2023; Sánchez-Prieto et al., 2019).

Finally, factors such as Individual-Level Cultural Values, Age, Gender, and Technological Innovation also moderate the adoption of AI tools, reflecting broader societal attitudes and individual differences that shape teachers’ willingness to embrace new technologies (Sánchez-Prieto et al., 2019; Zhang et al., 2023; Chocarro et al., 2021; Al Darayseh, 2023). Cultural values can influence perceptions of technology’s role in education, while demographic variables such as age and gender may affect comfort levels and attitudes toward AI (Alhumaid et al., 2023; Wang et al., 2021; Cukurova et al., 2023). The perception of technological innovation can encourage adoption if AI tools are seen as cutting-edge solutions that enhance teaching or discourage it if viewed as disruptive to established practices (Choi et al., 2022; Nazaretsky et al., 2021; Rico-Bautista et al., 2021).

Below is a visual that organizes the amount of evidence (number of studies) that supports a TAM variable’s importance when it comes to AI adoption as the dependent variable as opposed to general technology adoption.

Figure 3 – TAM Factors AI Adoption Evidence

Synthesis

To make sense of the large amount of information gathered, the researcher attempted to estimate the strength of influence of each of the factors given their outcome: General EdTech adoption or AI-Based EdTech adoption. While there is much more evidence for these factors in general EdTech adoption, the research took into consideration both the amount of literature available on AI adoption and the length of time the tools have been available to perform research and evaluation. Below is a figure (Figure 4) comparing the two types of adoption predictors and their estimated strengths, which were calculated by AI as described earlier.

Figure 4 – GenEdTech vs AI-Based EdTech Adoption Factors

Key Comparisons and Contrasts

• Self-efficacy:
  • A critical factor for both general EdTech and AI-based EdTech. Confidence in using technology is crucial across both types.
• Subjective Norms:
  • More influential for general EdTech. Social influence is less significant for AI-based tools, where personal judgment plays a bigger role.
• Perceived Enjoyment and Perceived Playfulness:
  • More impactful for general EdTech, indicating that enjoyment and intrinsic motivation are key for adopting traditional technologies, whereas AI tools are viewed more for their practical utility.
• System Quality and System Accessibility:
  • Important for both general and AI-based EdTech. The quality and ease of access to technology are consistently significant factors.
• Technological Complexity:
  • A major barrier for both, but particularly pronounced for AI-based EdTech (tied for 2nd strongest influencer in AI adoption with Technological Innovation as opposed to its 4-way tie for 2nd strongest influencer in Gen EdTech adoption) due to the perceived advanced nature of many of these tools according to the literature.
    • The strength of Technological Complexity and Technological Innovation are discussed to be very much related across multiple studies regarding AI adoption because they are concerned with the perception of the technology in many cases.
• Facilitating Conditions:
  • Essential for both types of technologies. The availability of resources and support strongly influences adoption.
• Anxiety:
  • Significant for both, with a higher impact on AI-based EdTech due to the perceived complexity and novelty associated with AI tools.
• Moderating Factors (Age, Gender, Cultural Values, Technological Innovation):
  • These factors have a moderate influence across both general and AI-based EdTech, with innovation perceived as slightly more influential for AI tools due to the large impact gen AI like LLms have had in education and the world in general.

Model

As a companion to the synthesis, a visual model was constructed to help explain the relationships between predictive factors and AI adoption. It depicts the strengths of relationships with lines of varying thicknesses, with the thick and more pronounced lines indicating stronger relationships and vice versa.

Key Components in the Diagram

1. Legend and Symbols:
   • Blue Gradient Bar (Top Left): Indicates the importance of each factor, with darker colors representing more important factors.
   • Arrows:
     • Thick Arrows: Major relationships that strongly influence other components.
     • Thin Arrows: Minor relationships that have a weaker influence.
   • Shapes:
     • Rounded Rectangles: Different types of factors affecting AI adoption.
     • Blue Ellipses and Circles: Non-TAM factors (not part of the original Technology Acceptance Model) that impact the process.
     • Yellow and Light Blue Rectangles: TAM factors that are central to the Technology Acceptance Model.
     • Green Rectangle: Behavioral intention, representing the intent to adopt AI.
     • Hexagon: Final AI adoption outcome.
2. TAM Factors (Core Factors in Technology Acceptance Model):
   • Perceived Ease of Use: Reflects the user’s belief that using AI will be free of effort.
   • Perceived Usefulness: Refers to the belief that AI will enhance the user’s effectiveness or job performance.
3. Non-TAM Factors (Additional Factors in the Extended Model):
   • System Quality: The overall performance and reliability of the AI system, which affects both ease of use and usefulness.
   • Technology Complexity: How difficult or complex the AI technology is to understand and use, impacting perceived ease of use.
   • Anxiety: Users’ level of apprehension or discomfort with AI, influencing perceived ease of use.
   • Self-Efficacy: The user’s confidence in their ability to use AI effectively.
   • System Accessibility: How easily users can access and engage with the AI technology, impacting perceived ease of use.
   • Perceived Enjoyment & Playfulness: Users’ perception of AI as enjoyable or fun, affecting their willingness to use it.
4. Additional Contextual Factors:
   • Facilitating Conditions: External support, resources, or infrastructure that help users adopt AI.
   • Subjective Norms: Social pressures or expectations that influence users’ perceived ease of use and usefulness.
   • Ethical Considerations: Ethical concerns regarding AI use, which influence the perceived usefulness and acceptance of AI.
   • Pedagogical Shift: Changes in teaching or training approaches due to AI, impacting perceived usefulness.
   • Cost & Time: Resources required for adopting AI, which influence perceived usefulness.
5. Moderators:
   • Age, Gender, Technological Innovation, and Cultural Values: Factors that modify the effect of ease of use and usefulness on behavioral intention. These variables help explain differences in AI adoption across demographic or cultural groups.

Flow and Relationships

• Direct Influences on Perceived Ease of Use:
  • System Quality, Technology Complexity, Anxiety, Self-Efficacy, and System Accessibility directly impact users’ perception of how easy AI is to use.
  • Facilitating Conditions and Subjective Norms also play a role, though they are more indirect and have a minor influence.
• Direct Influences on Perceived Usefulness:
  • Factors such as Ethical Considerations, Pedagogical Shift, Cost & Time, Subjective Norms, and Perceived Enjoyment & Playfulness contribute to how useful the AI is perceived.
  • System Quality and Technology Complexity impact usefulness as well, emphasizing the importance of a well-designed system that isn’t overly complex.
• Influences on Behavioral Intention:
  • Perceived Ease of Use and Perceived Usefulness are the two primary drivers of behavioral intention. They influence whether users intend to adopt AI.
  • Moderators such as age, gender, technological innovation, and cultural values adjust the impact of ease of use and usefulness on behavioral intention, recognizing that these factors are not universally felt in the same way by all users.
• Final Outcome – AI Adoption:
  • Behavioral Intention ultimately leads to AI adoption, where a higher intention to use AI translates to a greater likelihood of adoption.

Implications and Recommendations for AI Adoption in Education

The adoption of AI-based educational technology (EdTech) shares several predictors with general EdTech adoption but also presents unique challenges that require additional considerations. Understanding these similarities and differences has significant implications for educators, policymakers, and developers working to integrate AI into educational settings.

Similarities between AI Adoption and General EdTech Adoption

1. Perceived Ease of Use (PEU) and Perceived Usefulness (PU)
   Both AI and general EdTech adoption are heavily influenced by teachers’ perceptions of how easy the technology is to use (PEU) and how useful it is for enhancing teaching and learning (PU) (Davis, 1989). Factors like self-efficacy, system quality, and facilitating conditions play critical roles in shaping these perceptions (Scherer, Siddiq, & Tondeur, 2019; Venkatesh et al., 2003).
2. Behavioral Intention (BI)
   Self-efficacy and system accessibility are crucial in predicting BI, which represents the teacher’s intention to adopt the technology—an aspect essential for any technology adoption in educational contexts (Scherer, Siddiq, & Tondeur, 2019). Similarly, Fathema, Shannon, & Ross (2015) highlight that behavioral intention in educational settings is driven by the teachers’ confidence and access to the technology.
3. Moderating Factors
   Age, gender, and cultural values also moderately influence both AI and general EdTech adoption, suggesting that demographic and cultural factors shape how different groups perceive and adopt technology (Tarhini, Hone, & Liu, 2014; Sánchez-Prieto et al., 2020). These moderating factors help explain diverse adoption rates and attitudes among teachers.

Differences between AI Adoption and General EdTech Adoption

1. Technological Complexity and Anxiety
   AI tools are often perceived as more complex than general EdTech, presenting a higher barrier to adoption, which requires a more significant focus on training and support to enhance teachers’ confidence and ability to use these tools effectively (Howard, 2013; Nazaretsky, Cukurova, & Alexandron, 2021). Anxiety frequently accompanies perceived complexity, particularly with AI tools, as teachers may fear engaging with technologies they do not fully understand (Celik & Yesilyurt, 2013).
2. Ethical Concerns and Transparency
   AI adoption introduces ethical concerns, such as data privacy, algorithmic bias, and transparency in decision-making processes, which are less prevalent in general EdTech (Selwyn, 2019; Holmes et al., 2022). Addressing these issues is crucial for building trust and ensuring responsible AI use in education (Chocarro, Cortiñas, & Marcos-Matás, 2021).
3. Cost and Time
   The perceived cost and time associated with adopting AI tools are often higher than with general EdTech. Implementing AI may require significant investment in hardware, software, and professional development, adding a burden that schools must address to facilitate adoption (Woodruff, Hutson, & Arnone, 2023). Such concerns echo Cuban’s (1986) findings on the historical costs associated with adopting educational technology.
4. Required Shift in Pedagogy
   Unlike general EdTech, which can often be integrated with minimal changes to teaching methods, AI tools may necessitate significant shifts in pedagogy. Educators may need to rethink instructional strategies, assessments, and classroom management techniques to effectively incorporate AI into their practices (Trust et al., 2023; Zhang et al., 2023).

Implications for Educators and Policymakers

Given these similarities and differences, several key implications arise for those involved in integrating AI into education. It is essential to note that these actions primarily depend on support from education leaders and administrators, as many are beyond the scope of what a classroom teacher can accomplish.

1. Need for Comprehensive Training and Support
   Training programs must go beyond basic technology skills to include in-depth knowledge of AI tools, emphasizing the development of self-efficacy and reducing anxiety related to technological complexity (Ertmer, 1999). Alhumaid et al. (2023) also stress the importance of targeted training in building teachers’ confidence with AI tools.
2. Focus on Ethical Use and Transparency
   Educational institutions must develop guidelines on data privacy, algorithmic transparency, and equitable AI use (Selwyn, 2011; Al Darayseh, 2023). Transparent communication about how AI systems work and their ethical implications is fundamental to fostering trust and encouraging adoption.
3. Consideration of Cost and Time Investments
   Policymakers and school administrators should consider the higher costs and time investments associated with AI adoption. Providing financial support for purchases, as well as allocating time for professional development, can alleviate these barriers (Zhai et al., 2021; Buabeng-Andoh, 2012).
4. Support for Pedagogical Shifts
   As AI tools may require changes in teaching methods, there should be support structures in place to help teachers adapt their pedagogy. Providing resources, exemplars, and collaborative opportunities will enable educators to explore new instructional strategies enabled by AI technologies (Zawacki-Richter et al., 2019; Chuang, Shih, & Cheng, 2020).

Recommendations for Promoting AI Adoption

Based on the aforementioned implications, the following recommendations are proposed to promote AI adoption in educational settings. Strong support from district leaders and administrators will be critical, as subjective norms and facilitating conditions depend heavily on institutional support. Below in Table 2 are organized recommendations paired with resources followed by a detailed breakdown of each.

Table 2 – Recommendations

1. Develop Targeted and Comprehensive Training Programs
   Professional development should not only enhance teachers’ technical skills but also address unique aspects of AI, such as ethical considerations and pedagogical integration. Tailoring these programs to different demographic groups can address varying levels of comfort with technology (Ertmer, 1999; Alhumaid et al., 2023).
2. Simplify AI Tools and Ensure Usability
   Collaboration with developers to create intuitive, user-friendly AI tools, along with usability guides, can reduce technological complexity (Nazaretsky, Cukurova, & Alexandron, 2021).
3. Strengthen Institutional Support and Facilitate Access
   Robust institutional support through resources, technical assistance, and community-building channels allows teachers to share experiences and learn collaboratively (Woodruff, Hutson, & Arnone, 2023).
4. Promote Ethical Awareness and Transparency
   Clear guidelines on AI ethics are crucial. Workshops and discussions on ethical use will help build trust and ensure teachers understand the broader implications of AI in their classrooms (Selwyn, 2011; Al Darayseh, 2023).
5. Highlight Practical Benefits and Encourage Innovation
   Large-scale studies, case studies, and personal stories that showcase AI’s benefits in enhancing learning and teaching efficiency can inspire adoption and demonstrate AI’s value in solving educational challenges (Zhang et al., 2023; Choi, Jang, & Kim, 2022).

Limitations and Transparency

This meta-analysis was conducted with extensive use of AI tools, as discussed in the methodology section, but not without professional researcher supervision and checks for accuracy. This approach marks a departure from traditional meta-analytic studies, which often adhere to specific, manual procedures for analyzing and selecting studies (Gough, Oliver, & Thomas, 2017). In contrast, this study leveraged artificial intelligence, specifically large language models (LLMs), to streamline data categorization, sorting, and initial analytics.

While these methods diverge from traditional practices, they are not without merit or justification. Hullman (2024) emphasizes the practical and accurate potential of using LLMs as tools for categorization and data analysis across diverse datasets. ChatGPT operated in this study as a rule-driven assistant, categorizing, sorting, and performing preliminary analytics to deliver digestible results, which the researcher then used for comparative analysis across categories.

Research supports the idea that AI can perform categorization tasks with near-human accuracy. For instance, Khraisha, Put, Kappenberg, Warraitch, and Hadfield (2024) found that while humans generally outperform AI in meta-analytic procedures, LLMs are highly capable of following structured instructions for categorization and basic analytics, achieving “almost perfect performance on par with humans.” Similarly, studies by Shank and Wilson (2023) and Chelli et al. (2024) discuss the strengths of AI in handling large datasets when guided by specific parameters.

Conversely, some studies (Chelli et al., 2024; Cheloff, 2023) indicate limitations in using LLMs exclusively for systematic literature reviews due to issues with accuracy and recall. This study, however, mitigated these concerns by using LLMs not as exclusive tools but as aids in finding and sorting studies under researcher-defined rules, with all outputs cross-verified for accuracy (Lewis et al., 2023).

Ultimately, while AI in research and analytics is still in its early stages and has limitations, it also presents advantages, offering efficiency and accuracy comparable to human performance when used with appropriate guidance (Hancock & Beebe, 2023; Song, 2024).

Conclusions

This meta-analysis confirms the persistence of key factors influencing educators’ adoption of technology, with specific emphasis on AI in educational settings. As noted in the introduction, resistance to technology in education—from calculators to computers—has followed a familiar trajectory, and AI is no exception. This study, building on foundational models like the Technology Acceptance Model (TAM) and extending the work of Scherer, Siddiq, and Tondeur (2019) and Grannic (2022), highlights that Perceived Ease of Use (PEU) and Perceived Usefulness (PU) continue to be central determinants in teachers’ acceptance of AI technologies. However, AI introduces complexities beyond those seen with previous technologies.

In particular, critical factors such as Self-Efficacy, Cost & Time, and the Required Pedagogical Shift emerged as highly influential in this meta-analysis. Self-efficacy, or teachers’ confidence in their ability to use AI tools, is a significant predictor of adoption, underscoring the need for targeted professional development and support. This echoes Ertmer’s (1999) findings about the importance of training in overcoming technological resistance. Cost & Time, often overlooked in broader technology adoption discussions, play a more pronounced role in AI adoption, as teachers perceive AI tools to require significant financial investment and time to learn. This aligns with Zhai et al. (2021) findings that teachers feel unprepared and overwhelmed by the demands of new technology, a barrier that could hinder adoption unless schools provide resources and time for teachers to adapt.

The Required Pedagogical Shift further complicates AI adoption. Unlike previous technologies that have been integrated with extensive support to help teachers understand how to utilize the tools effectively in their current pedagogical contexts, AI tools often necessitate a fundamental rethinking of teaching strategies and classroom management. Teachers may resist adopting AI because it challenges traditional teaching practices, an issue identified in both this meta-analysis and by Trust et al. (2023), who noted similar concerns about the impact of AI on instructional methods and critical thinking skills. Overcoming this resistance will require not just technical training but support for pedagogical innovation.

Anxiety, a recurring theme in technology adoption, was another strong predictor of resistance. Educators expressed significant apprehension about the complexity and perceived risks of AI, particularly in relation to job displacement and the ethical challenges of AI in education, such as data privacy and bias. This anxiety, as reported by Holmes et al. (2022) and highlighted in our findings, must be addressed by increasing transparency, providing ethical guidelines, and offering continuous support to educators as they integrate AI into their practices.

Beyond these core factors, System Accessibility and Technological Complexity remain central barriers to AI adoption as teachers struggle with the perceived difficulty of navigating AI tools. This echoes historical patterns of resistance, as seen in Cuban (1986)’s documentation of educators’ reluctance to adopt past technologies. Overcoming these obstacles, much like with earlier innovations, will depend on creating user-friendly systems and reducing the perceived burden on teachers.

While addressing these factors is crucial, it is also essential to acknowledge the value of resistance to technology adoption, as highlighted by The Friction Project (Sutton & Rao, 2024), and Noise: A flaw in human judgment, which argues that friction can serve as a healthy barrier, encouraging thoughtful consideration and critical questioning of new tools before widespread adoption (Kahneman, Sibony & Sunstein, 2021). In this context, resistance to AI may signal legitimate concerns that should be addressed rather than dismissed. For instance, educators’ reluctance to adopt AI may stem from valid ethical considerations, such as privacy concerns and the impact on students’ critical thinking skills. Recognizing these concerns and addressing them thoughtfully can ensure that adoption is more purposeful and aligned with educational goals, rather than simply following technological trends.

In conclusion, this meta-analysis identifies a set of predictors—PEU, PU, Self-Efficacy, Anxiety, Cost & Time, and Required Pedagogical Shift—that both align with historical trends and reveal new challenges unique to AI. While core principles of technology acceptance, such as Perceived Usefulness and Self-Efficacy, continue to shape adoption, the high demands of AI in terms of time, cost, and pedagogical adjustment suggest that this latest technological wave requires more comprehensive and tailored support than its predecessors. Addressing these factors, along with mitigating ethical concerns, reducing anxiety, and recognizing the constructive role of resistance, will be essential for overcoming barriers and ensuring that AI can effectively enhance educational practices. This approach aligns with the ideas of Piaget (1936) and Bransford et al. (2000), who advocated for thoughtful integration of technology to foster meaningful learning experiences while also taking into account the reasoning for friction in AI adoption and addressing it.

References

Agarwal, R., & Prasad, J. (1998). A conceptual and operational definition of personal innovativeness in the domain of information technology. Information Systems Research, 9(2), 204-215.

Al Darayseh, A. (2023). Acceptance of artificial intelligence in teaching science: Science teachers’ perspective. Computers and Education: Artificial Intelligence, 4, 100132.

Alhumaid, K., Ali, S., Waheed, A., Zahid, E., & Habes, M. (2021). COVID-19 & elearning: Perceptions & attitudes of teachers towards e-learning acceptance in the developing countries. Multicultural Education, 7(2), 100-115.

Alhumaid, K., Naqbi, S., Elsori, D., & Mansoori, M. (2023). The adoption of artificial intelligence applications in education. International Journal of Data and Network Science, 7(1), 457-466.

Aldunate, R., & Nussbaum, M. (2013). Teacher adoption of technology. Computers in Human Behavior, 29(3), 519-524.

Ayanwale, M. A., Sanusi, I. T., Adelana, O. P., Aruleba, K. D., & Oyelere, S. S. (2022). Teachers’ readiness and intention to teach artificial intelligence in schools. Computers and Education: Artificial Intelligence, 3, 100099.

Bransford, J. D., Brown, A. L., & Cocking, R. R. (Eds.). (2000). How people learn: Brain, mind, experience, and school (Expanded ed.). National Academy Press.

Buabeng-Andoh, C. (2012). Factors influencing teachers’ adoption and integration of information and communication technology into teaching: A review of the literature. International Journal of Education and Development using Information and Communication Technology, 8(1), 136-155.

Calisir, F., Altin Gumussoy, C., Bayraktaroglu, A. E., & Karaali, D. (2014). Predicting the intention to use a web‐based learning system: Perceived content quality, anxiety, perceived system quality, image, and the technology acceptance model. Human Factors and Ergonomics in Manufacturing & Service Industries, 24(5), 515-531.

Celik, V., & Yesilyurt, E. (2013). Attitudes to technology, perceived computer self-efficacy and computer anxiety as predictors of computer supported education. Computers & Education, 60(1), 148-158.

Chatterjee, S., & Bhattacharjee, K. K. (2020). Adoption of artificial intelligence in higher education: A quantitative analysis using structural equation modelling. Education and Information Technologies, 25, 3443-3463.

Chelli, A., Collins, T., Darwish, M., Fan, Y., & Kraus, M. (2024). The limitations of AI in literature reviews: A comparative study. Journal of Digital Scholarship, 11(2), 45-62.

Chelli, M., Descamps, J., Lavoué, V., Trojani, C., Azar, M., Deckert, M., … & Ruetsch-Chelli, C. (2024). Hallucination Rates and Reference Accuracy of ChatGPT and Bard for Systematic Reviews: Comparative Analysis. Journal of Medical Internet Research, 26, e53164.

Cheloff, S. (2023). AI and systematic reviews: Challenges and considerations. New York, NY: Research Press.

Cheloff, A. Z., Pochapin, M., & Popov, V. (2023). S1726 Publicly Available Generative Artificial Intelligence Programs Are Currently Unsuitable for Performing Meta-Analyses. Official journal of the American College of Gastroenterology| ACG, 118(10S), S1287.

Chen, I. J., Yang, K. F., Tang, F. I., Huang, C. H., & Yu, S. (2008). Applying the technology acceptance model to explore public health nurses’ intentions towards web-based learning: A cross-sectional questionnaire survey. International journal of nursing studies, 45(6), 869-878.

Cheung, R., & Vogel, D. (2013). Predicting user acceptance of collaborative technologies: An extension of the technology acceptance model for e-learning. Computers & education, 63, 160-175.

Chang, C. T., Hajiyev, J., & Su, C. R. (2017). Examining the students’ behavioral intention to use e-learning in Azerbaijan? The general extended technology acceptance model for e-learning approach. Computers & Education, 111, 128-143.

Cheng, Y. M. (2019). How does task-technology fit influence cloud-based e-learning continuance and impact? Education + Training, 61(4), 480-499.

Chiu, T. K. (2021). Applying the self-determination theory (SDT) to explain student engagement in online learning during the COVID-19 pandemic. Journal of Research on Technology in Education, 54(1), 14-30.

Chocarro, R., Cortiñas, M., & Marcos-Matás, G. (2021). Teachers’ attitudes towards chatbots in education: a technology acceptance model approach considering the effect of social language, bot proactiveness, and users’ characteristics. Educational Studies, 49, 295 – 313.

Choi, S.Y., Jang, Y., & Kim, H. (2022). Influence of Pedagogical Beliefs and Perceived Trust on Teachers’ Acceptance of Educational Artificial Intelligence Tools. International Journal of Human–Computer Interaction, 39, 910 – 922.

Chuang, H. H., Shih, C. L., & Cheng, M. M. (2020). Teachers’ perceptions of culturally responsive teaching in technology-supported learning environments. British Journal of Educational Technology, 51(6), 2442-2460.

Cuban, L. (1986). Teachers and machines: The classroom use of technology since 1920. Teachers College Press.

Cukurova, M., Miao, X., & Brooker, R. (2023, June). Adoption of artificial intelligence in schools: unveiling factors influencing teachers’ engagement. In International conference on artificial intelligence in education (pp. 151-163). Cham: Springer Nature Switzerland.

Darayseh, A.A. (2023). Acceptance of artificial intelligence in teaching science: Science teachers’ perspective. Comput. Educ. Artif. Intell., 4, 100132.

Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319-340.

Ertmer, P. A. (1999). Addressing first- and second-order barriers to change: Strategies for technology integration. Educational Technology Research and Development, 47(4), 47-61.

Fathema, N., Shannon, D., & Ross, M. (2015). Expanding the Technology Acceptance Model (TAM) to examine faculty use of Learning Management Systems (LMSs) in higher education institutions. Journal of Online Learning & Teaching, 11(2), 210-232.

Gough, D., Oliver, S., & Thomas, J. (2017). An introduction to systematic reviews. London, UK: SAGE Publications.

Granić A. (2022). Educational Technology Adoption: A systematic review. Education and information technologies, 27(7), 9725–9744. https://doi.org/10.1007/s10639-022-10951-7

Guillén-Gámez, F. D., & Mayorga-Fernández, M. J. (2020). Identification of variables that predict teachers’ attitudes toward ICT in higher education for teaching and research: A study with regression. Sustainability, 12(4), 1312.

Hancock, R., & Beebe, S. (2023). The impact of large language models on data analytics: A study of efficiency and accuracy. Journal of Data Science and AI, 6(1), 29-47.

Hanif, A., Jamal, F. Q., & Imran, M. (2018). Extending the technology acceptance model for use of e-learning systems by digital learners. Ieee Access, 6, 73395-73404.

Holden, H., & Rada, R. (2011). Understanding the influence of perceived usability and technology self-efficacy on teachers’ technology acceptance. Journal of Research on Technology in Education, 43(4), 343-367.

Howard, S. K. (2013). Risk-aversion: Understanding teachers’ resistance to technology integration. Technology, Pedagogy and Education, 22(3), 357-372.

Hsu, H. H., & Chang, Y. Y. (2013). Extended TAM model: Impacts of convenience on acceptance and use of Moodle. US-China Education Review, 3(4), 211-218.

Hsu, L. (2020). Factors affecting adoption of digital teaching in elementary school English: A mixed methods study. Computer Assisted Language Learning, 1-23.

Hullman, J. (2024). Practical applications of LLMs in data categorization and analysis. Journal of Computational Methods, 8(1), 12-25.

Hullman, J. (2024, June 24). Forking paths in LLMs for data analysis. Statistical Modeling, Causal Inference, and Social Science. https://statmodeling.stat.columbia.edu/2024/06/24/forking-paths-in-llms-for-data-analysis/

Joo, Y. J., Park, S., & Lim, E. (2018). Factors influencing preservice teachers’ intention to use technology: TPACK, teacher self-efficacy, and technology acceptance model. Educational Technology & Society, 21(3), 48-59.

Kahneman, D., Sibony, O., & Sunstein, C. R. (2021). Noise: A flaw in human judgment. New York: Little, Brown Spark.

Khraisha, Q., Put, S., Kappenberg, J., Warraitch, A., & Hadfield, K. (2024). Can large language models replace humans in systematic reviews? Evaluating GPT‐4’s efficacy in screening and extracting data from peer‐reviewed and grey literature in multiple languages. Research Synthesis Methods.

König, J., Jäger-Biela, D. J., & Glutsch, N. (2020). Adapting to online teaching during COVID-19 school closure: Teacher education and teacher competence effects among early career teachers in Germany. European Journal of Teacher Education, 43(4), 608-622.

Lawrence, J. E., & Tar, U. A. (2018). Factors that influence teachers’ adoption and integration of ICT in teaching/learning process. Educational Media International, 55(1), 79-105.

Leem, J., & Sung, E. (2019). Teachers’ beliefs and technology acceptance concerning smart mobile devices for SMART education in South Korea. British Journal of Educational Technology, 50(2), 601-613.

Lewis, C., Martinez, F., Olson, H., & Chen, L. (2023). Ensuring accuracy in AI-assisted systematic reviews: A mixed-methods approach. International Journal of AI in Research, 2(4), 205-223.

Liu, H., Wang, L., & Koehler, M. J. (2019). Exploring the intention-behavior gap in the technology acceptance model: A mixed-methods study in the context of foreign-language teaching in China. British Journal of Educational Technology, 50(5), 2536-2556.

Mac Callum, K., Jeffrey, L., & Kinshuk. (2014). Factors impacting teachers’ adoption of mobile learning. Journal of Information Technology Education: Research, 13, 141-162.

Mailizar, M., Burg, D., & Maulina, S. (2021). Examining university teachers’ acceptance of learning management system (LMS) in Indonesia: A rasch analysis approach. Education and Information Technologies, 26(4), 4089-4108.

McGehee, N. (2023) Balancing the Risks and Rewards of AI Integration for Michigan Teachers. Michigan Virtual. https://michiganvirtual.org/research/publications/balancing-the-risks-and-rewards-of-ai-integration-for-michigan-teachers/

McGehee, N. (2024, June 21). AI in education: Student usage in online learning. Michigan Virtual Learning Research Institute. https://michiganvirtual.org/research/publications/ai-in-education-student-usage-in-online-learning/

Michigan Virtual. (2024). AI in Education: Exploring Trust, Challenges, and the Push for Implementation. https://michiganvirtual/research/publications/ai-in-education-exploring-trust-challenges-and-the-push-for-implementation/

Mohammadi, H. (2015). Investigating users’ perspectives on e-learning: An integration of TAM and IS success model. Computers in Human Behavior, 45, 359-374.

Moran, M., Hawkes, M., & Gayar, O. E. (2010). Tablet personal computer integration in higher education: Applying the unified theory of acceptance and use technology model to understand supporting factors. Journal of educational computing research, 42(1), 79-101.

Muhaimin, M., Habibi, A., Mukminin, A., Saudagar, F., Pratama, R., Wahyuni, S., … & Indrayana, B. (2019). A sequential explanatory investigation of TPACK: Indonesian science teachers’ survey and perspective. Journal of Technology and Science Education, 9(3), 269-281.

Mun, Y. Y., & Hwang, Y. (2003). Predicting the use of web-based information systems: self-efficacy, enjoyment, learning goal orientation, and the technology acceptance model. International journal of human-computer studies, 59(4), 431-449.

Nagy, J. T. (2018). Evaluation of online video usage and learning satisfaction: An extension of the technology acceptance model. International Review of Research in Open and Distributed Learning, 19(1).

Nam, C. S., Bahn, S., & Lee, R. (2013). Acceptance of assistive technology by special education teachers: A structural equation model approach. International Journal of Human-Computer Interaction, 29(5), 365-377.

Nazaretsky, T., Cukurova, M., & Alexandron, G. (2021). An Instrument for Measuring Teachers’ Trust in AI-Based Educational Technology. LAK22: 12th International Learning Analytics and Knowledge Conference.

Nistor, N., Göğüş, A., & Lerche, T. (2013). Educational technology acceptance across national and professional cultures: a European study. Educational Technology Research and Development, 61(4), 733-749.

Nja, C. O., Idiege, K. J., Uwe, U. E., Meremikwu, A. N., Ekon, E. E., Erim, C. M., … & Cornelius-Ukpepi, B. U. (2023). Adoption of artificial intelligence in science teaching: From the vantage point of the African science teachers. Smart Learning Environments, 10(1), 42.

O’Bannon, B. W., & Thomas, K. (2014). Teacher perceptions of using mobile phones in the classroom: Age matters! Computers & Education, 74, 15-25.

Padilla-Meléndez, A., del Aguila-Obra, A. R., & Garrido-Moreno, A. (2013). Perceived playfulness, gender differences and technology acceptance model in a blended learning scenario. Computers & Education, 63, 306-317.

Park, S. Y., Nam, M. W., & Cha, S. B. (2012). University students’ behavioral intention to use mobile learning: Evaluating the technology acceptance model. British journal of educational technology, 43(4), 592-605.

Piaget, J. (1936). Origins of intelligence in the child. Routledge & Kegan Paul.

Qasem, A. A. A., & Viswanathappa, G. (2020). The integration of multiliteracies in digital learning environments: A study of teacher readiness. International Journal of Technology in Education and Science, 4(4), 265-279.

Rico-Bautista, D., Medina-Cardenas, Y., Coronel-Rojas, L. A., Cuesta-Quintero, F., Maestre-Gongora, G., & Guerrero, C. D. (2021). Smart university: key factors for an artificial intelligence adoption model. In Advances and Applications in Computer Science, Electronics and Industrial Engineering: Proceedings of CSEI 2020 (pp. 153-166). Singapore: Springer Singapore.

Rogers, E. M. (2003). Diffusion of innovations (5th ed.). New York: Free Press.

Salloum, S. A., Alhamad, A. Q. M., Al-Emran, M., Monem, A. A., & Shaalan, K. (2019). Exploring students’ acceptance of e-learning through the development of a comprehensive technology acceptance model. IEEE access, 7, 128445-128462.

Sánchez-Prieto, J. C., Olmos-Migueláñez, S., & García-Peñalvo, F. J. (2017). MLearning and pre-service teachers: An assessment of the behavioral intention using an expanded TAM model. Computers in Human Behavior, 72, 644-654.

Sánchez-Prieto, J.C., Cruz-Benito, J., Therón, R., & García-Peñalvo, F.J. (2019). How to Measure Teachers’ Acceptance of AI-driven Assessment in eLearning: A TAM-based Proposal. Proceedings of the Seventh International Conference on Technological Ecosystems for Enhancing Multiculturality.

Sánchez-Prieto, J. C., Huang, F., Olmos-Migueláñez, S., García-Peñalvo, F. J., & Teo, T. (2020). Exploring the unknown: The effect of resistance to change and attachment on mobile adoption among secondary pre-service teachers. British Journal of Educational Technology, 51(3), 626-643.

Sánchez-Cruzado, C., Santiago Campión, R., & Sánchez-Compaña, M. T. (2021). Teacher digital literacy: The indisputable challenge after COVID-19. Sustainability, 13(4), 1858.

Scherer, R., Siddiq, F., & Tondeur, J. (2019). The technology acceptance model (TAM): A meta-analytic structural equation modeling approach to explaining teachers’ adoption of digital technology in education. Computers & Education, 128, 13-35.

Scherer, R., & Teo, T. (2019). Unpacking teachers’ intentions to integrate technology: A meta-analysis. Educational Research Review, 27, 90-109.

Scherer, R., Howard, S. K., Tondeur, J., & Siddiq, F. (2021). Profiling teachers’ readiness for online teaching and learning in higher education: Who’s ready? Computers in Human Behavior, 118, 106675.

Selwyn, N. (2011). Education and technology: Key issues and debates. Continuum International Publishing Group.

Shank, M., & Wilson, J. (2023). AI’s role in data-intensive research: From sorting to summarizing. Washington, DC: Academic Research Institute.

Song, L. (2024). Efficiency and reliability of AI in research synthesis. Journal of Advanced Research Methods, 9(2), 34-50.

Song, Y., & Kong, S. C. (2017). Investigating students’ acceptance of a statistics learning platform using technology acceptance model. Journal of Educational Computing Research, 55(6), 865-897.

Sutton, R. I., & Rao, H. (2024). The friction project: How smart leaders make the right things easier and the wrong things harder. St. Martin’s Press.

Tarhini, A., Hone, K., & Liu, X. (2014). Measuring the moderating effect of gender and age on e-learning acceptance in England: A structural equation modeling approach for an extended technology acceptance model. Journal of Educational Computing Research, 51(2), 163-184.

Teo, T. (2009). Modeling technology acceptance in education: A study of pre-service teachers. Computers & Education, 52(2), 302-312.

Teo, T. (2010). A path analysis of pre-service teachers’ attitudes to computer use: applying and extending the technology acceptance model in an educational context. Interactive Learning Environments, 18(1), 65-79.

Teo, T. (2011). Factors influencing teachers’ intention to use technology: Model development and test. Computers & Education, 57(4), 2432-2440.

Teo, T., & Noyes, J. (2008). Development and validation of a computer attitude measure for young students (CAMYS). Computers in Human Behavior, 24(6), 2659-2667.

Teo, T., & Noyes, J. (2011). An assessment of the influence of perceived enjoyment and attitude on the intention to use technology among pre-service teachers: A structural equation modeling approach. Computers & Education, 57(2), 1645-1653.

Teo, T., & van Schaik, P. (2012). Understanding the intention to use technology by preservice teachers: An empirical test of competing theoretical models. International Journal of Human-Computer Interaction, 28(3), 178-188.

Teo, T., Luan, W. S., & Sing, C. C. (2008). A cross-cultural examination of the intention to use technology between Singaporean and Malaysian pre-service teachers: an application of the Technology Acceptance Model (TAM). Educational Technology & Society, 11(4), 265-280.

Teo, T., Huang, F., & Hoi, C. K. W. (2018). Explicating the influences that explain intention to use technology among English teachers in China. Interactive Learning Environments, 26(4), 460-475.

Teo, T., Zhou, M., & Noyes, J. (2016). Teachers and technology: Development of an extended theory of planned behavior. Educational Technology Research and Development, 64(6), 1033-1052.

Teo, T., Huang, F., & Hoi, C. K. W. (2019). Towards a new model of teachers’ integration of technology: A conceptual framework. British Journal of Educational Technology, 50(5), 2476-2493.

Teo, T., Huang, F., & Hoi, C. K. W. (2021). Explicating the influences that explain intention to use technology among English teachers in China. Interactive Learning Environments, 29(3), 429-443.