DRL scaffolding significantly benefits high prior knowledge students while BKT does not, and condition-by-prior-knowledge interactions are not significant¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r2 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? High-DRL students achieved a significantly higher posttest score than High-Control students (β= 6.81, p= 0.008), while High-BKT did not (p=.16). → Tithi 2026 q3 i? The condition-by-prior-knowledge interaction terms were not statistically significant, indicating the adaptive advantage did not differ significantly by prior knowledge level. → Tithi 2026 (2)

Evidence¶

Tithi 2026¶

Tithi, S.D., Alam, N., Yasir, T., Shi, Y., Tian, X., Chi, M., Barnes, T. (2026). Adaptive Scaffolding for Cognitive Engagement in an Intelligent Tutoring System. arXiv. https://arxiv.org/abs/2602.07308 doi:10.48550/arxiv.2602.07308

q3 · i? · causal · r2

Follow-up pairwise comparisons within the High prior knowledge subgroup of the N=113 mixed-effects regression (High-Control n=18, High-BKT n=20, High-DRL n=20), examining condition differences within the High subgroup.

"studentsinHigh-DRLachievedasignificantlyhigherposttestscorethanstudents in High-Control (β= 6.81,SE= 2.6,z= 2.6,p= 0.008), while students in High-BKT did not (β= 4.09,SE= 2.9,z= 1.4,p=.16)."

Tithi 2026 (2)¶

Tithi, S.D., Alam, N., Yasir, T., Shi, Y., Tian, X., Chi, M., Barnes, T. (2026). Adaptive Scaffolding for Cognitive Engagement in an Intelligent Tutoring System. arXiv. https://arxiv.org/abs/2602.07308. doi:10.48550/arxiv.2602.07308

q3 · i? · causal · r2

Interaction terms from the same mixed-effects regression model on posttest problem scores testing whether the adaptive advantage over Control varied by prior knowledge level; neither interaction reached significance.

"Theinteractiontermswerenotstatisticallysignificant(High-BKT:β=−5.3, SE= 3.9,z=−1.4,p=.18; High-DRL:β= 2.0,SE= 3.8,z= 0.6,p= 0.59), indicating that the advantage of adaptive conditions overControldid not differ significantly by prior knowledge level."

Discussion¶

Learner Variables¶

• Prior Knowledge — moderator: an instructional effect differs with it

Related Claims¶

• Instructional guidance that helps novices can become redundant or counterproductive as expertise grows. — the general claim this finding bears on, with the studies behind it
• Adaptive scaffolding policies (BKT and DRL) significantly improve posttest performance over a non-adaptive control in a logic ITS — a broader claim this one bears on
• BKT scaffolding significantly benefits low prior knowledge students, who outperformed low prior knowledge controls — related
• All training policies narrowed the prior-knowledge achievement gap from pretest to posttest, with BKT achieving the largest reduction (77.1%) — related
• The DRL policy produced a significantly different scaffolding distribution than Control and BKT, favoring Guided examples (60%) and avoiding Buggy examples (4%) — related
• No significant differences in posttest rule accuracy across scaffolding conditions — reports the opposite
• Critical thinking improvement differed by proficiency level: high-level students gained 18%, intermediate 12%, and low-level 8%, with asymmetric feedback adaptation — related
• The delayed erroneous-examples benefit holds for both low and high prior knowledge students — related
• Pretest accuracy moderated the training effect for typical full-fledged arguments: students with high pretest scores benefited most — related
• Study performance moderated the training effect: high-achieving students benefited at posttest, low performers did not — reports the opposite
• The effect of mathematics test practice varies across prior-achievement levels, with significant effects at intermediate-to-high levels (levels 3 and 4) but not at the lowest or highest — related
• Learning benefits of Cognitive Tutor interactivity are supported to varying extents by a growing body of experimental studies — related