Whether adaptive computer-assisted instruction (CAI) improves a specific reading-comprehension skill more than the same explicit instruction delivered by a teacher remains an open empirical question. A 2021 randomized controlled trial with 47 struggling upper-elementary readers found that the adaptive CAI program Digital Comprehension produced greater growth in main-idea identification than business-as-usual instruction, but its design could not separate the effect of the adaptive technological delivery from the effect of receiving additional explicit main-idea instruction, could not rule out a novelty effect, and assessed only immediate posttests. Later syntheses (2024–2026) of technology-delivered literacy instruction pool all comparison conditions and explicitly identify the technology-versus-equivalent-teacher-led comparison as an uncoded, open research need. The present problem asks for a randomized comparison of adaptive main-idea CAI against a content- and time-matched teacher-led intervention — with standardized measurement and immediate plus delayed posttests — that isolates the delivery effect and establishes whether gains persist in struggling readers.

Background

Computer-assisted instruction (CAI) is instruction in which a computer program, rather than a teacher, presents content, prompts responses, and delivers feedback. In adaptive CAI the program additionally adjusts task difficulty, sequence, or pacing in real time to an individual student's performance, a design closely related to intelligent tutoring systems, computer programs that tutor one learner at a time and adapt instruction to that learner's responses. Identifying the main idea of a text — stating its central gist rather than recalling details — is a foundational comprehension skill that struggling readers (students whose reading performance falls substantially below grade expectations) rarely acquire without explicit teaching, and main-idea strategy instruction is a common target of supplemental intervention. In a randomized controlled trial (RCT), participants are randomly assigned to conditions so that subsequent differences in outcomes can be attributed to the assigned condition rather than to pre-existing differences; a business-as-usual (BAU) control group receives whatever ordinary classroom instruction it would get, with nothing added.

Toonder and Sawyer (2021) reported a small RCT of a researcher-developed adaptive CAI program, Digital Comprehension, with 47 upper-elementary struggling readers. Treated students, who worked through the adaptive program while controls received only BAU instruction, showed significantly greater growth on a main-idea measure; no main effect appeared on a general comprehension measure. The authors' discussion — the source of the present problem — notes that this design cannot separate three contributions: the adaptive technological delivery itself, the benefit of simply receiving additional explicit main-idea instruction grounded in a validated model, and a novelty effect, the well-documented phenomenon in educational technology research in which the newness of a technologically delivered activity temporarily raises motivation and performance. Because only immediate posttests were administered, the durability of the gain was unknown, and the main-idea outcome was a researcher-developed rather than standardized instrument, limiting comparability across settings.

Adjacent literature informs but does not answer the question. VanLehn (2011) reviewed the relative effectiveness of human tutoring, intelligent tutoring systems, and other tutoring systems, a synthesis tradition showing that machine-delivered tutoring can approach human tutoring on average across domains; that literature does not concern struggling elementary readers' comprehension-skill remediation, nor does it control novelty or assess durability for this population. Meta-analytic syntheses of technology-delivered literacy instruction report small-to-moderate average advantages over comparison conditions — Dahl-Leonard, Hall, and Peacott (2024) estimate a mean Hedges' gg of 0.240.24 (a standardized mean difference, where 0.20.2 is conventionally small and 0.50.5 medium) across 53 studies for elementary literacy, and Alqahtani (2026) estimates g=0.35g = 0.35 across 30 RCTs for elementary students with reading difficulties — but both accept as comparison 'any group not receiving technology-delivered literacy instruction', pooling BAU and teacher-delivered contrasts. General meta-analyses of technology-enhanced adaptive learning (Wang, Tigelaar, Ye, and Admiraal, 2025, 69 studies) and of mobile-device integration (Sung, Chang, and Liu, 2016, 110 studies, mean effect 0.5230.523) likewise aggregate across heterogeneous comparison types. Dahl-Leonard et al. (2024) state explicitly that whether technology-delivered instruction complements or replaces teacher-led instruction was not coded and is needed future research. The unresolved target is therefore the delivery-isolated head-to-head contrast and its persistence over time.

Problem Statement

In struggling upper-elementary readers, does adaptive CAI targeting main-idea identification produce greater immediate and maintained gains in main-idea identification — and secondarily in general reading comprehension — than an equivalent teacher-led explicit main-idea intervention, when the delivery effect is isolated from the added-instruction and novelty confounds that limit the existing evidence?

That question concerns a randomized controlled trial that assigns struggling upper-elementary readers to (a) adaptive CAI for main-idea identification (of the Digital Comprehension type) or (b) a teacher-led explicit main-idea intervention matched to the CAI arm in instructional content and total instructional time; measures main-idea identification with a standardized or otherwise cross-setting comparable instrument, plus general reading comprehension, at pretest, immediate posttest, and at least one delayed posttest weeks to months after the intervention ends; controls the novelty effect; and reports the pairwise contrasts — above all the adaptive-CAI-versus-teacher-led difference at immediate and delayed posttests. A business-as-usual reference arm and exploratory moderator analyses (for example, for English learners or students receiving special education services) may accompany the comparison as secondary elements.

References

  1. Toonder, S., & Sawyer, L. B. (2021). The impact of adaptive computer assisted instruction on reading comprehension: Identifying the main idea. Journal of Computer Assisted Learning, 37(5), 1336–1347. DOI: 10.1111/jcal.12573. https://doi.org/10.1111/jcal.12573
  2. Dahl-Leonard, K., Hall, C., & Peacott, D. (2024). A meta-analysis of technology-delivered literacy instruction for elementary students. Educational Technology Research and Development, 72(3), 1507–1538. DOI: 10.1007/s11423-024-10354-0. https://doi.org/10.1007/s11423-024-10354-0
  3. Wang, J., Tigelaar, D. E. H., Ye, T., & Admiraal, W. (2025). A Meta-Analysis of Moderators of the Effects of Technology-Enhanced Adaptive Learning on Primary and Secondary Students' Learning Outcomes. Journal of Computer Assisted Learning, 42(1), Article e70168. DOI: 10.1002/jcal.70168. https://doi.org/10.1002/jcal.70168
  4. Alqahtani, S. S. (2026). A meta-analysis of randomized controlled trials on technology-based interventions for elementary students with reading difficulties. Research in Developmental Disabilities, 174, Article 105318. DOI: 10.1016/j.ridd.2026.105318. https://doi.org/10.1016/j.ridd.2026.105318
  5. Yan, X., Peng, P., & Liu, Y. (2024). Optimal design feature of computer-assisted reading instruction for students with reading difficulties? A Bayesian network meta-analysis. Computers in Human Behavior, 152, Article 108062. DOI: 10.1016/j.chb.2023.108062. https://doi.org/10.1016/j.chb.2023.108062
  6. Young, M. K., King, S. A., Datchuk, S. M., McFadden, K. A., & Yoon, H. (2026). Fidelity of Implementation and Instructional Alignment in Tier 2 Interventions: An Updated Review of Experimental Studies. Learning Disabilities Research & Practice. Advance online publication. DOI: 10.1177/09388982261476719. https://doi.org/10.1177/09388982261476719
  7. Butterfuss, R., McCarthy, K. S., Orcutt, E., Kendeou, P., & McNamara, D. S. (2024). Identification of main ideas in expository texts: selection versus deletion. Reading and Writing, 37(3), 757–785. DOI: 10.1007/s11145-023-10431-5. https://doi.org/10.1007/s11145-023-10431-5
  8. VanLehn, K. (2011). The relative effectiveness of human tutoring, intelligent tutoring systems, and other tutoring systems. Educational Psychologist, 46(4), 197–221. DOI: 10.1080/00461520.2011.611369. https://doi.org/10.1080/00461520.2011.611369
  9. Sung, Y.-T., Chang, K.-E., & Liu, T.-C. (2016). The effects of integrating mobile devices with teaching and learning on students' learning performance: A meta-analysis and research synthesis. Computers & Education, 94, 252–275. DOI: 10.1016/j.compedu.2015.11.008. https://doi.org/10.1016/j.compedu.2015.11.008