A contingent AI tutor places metacognitive demand equal to a question-only tutor and higher than an answer-on-request assistant¶

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? Under contingent support, learners reported more effort on the task and on the approach than under the answer-on-request tutor, and did not differ from the question-only tutor on approach effort. → Hou 2026 q3 i? Learners most often named the contingent tutor as the one they thought hardest with and did the most thinking themselves. → Hou 2026 (2)

Evidence¶

Hou 2026¶

Hou, X., Weng, Y., Yeh, C. H., Lee, D. L., Zheng, L., Li, F., Wang, W., & Liu, Y. (2026). Adaptive Scaffolding Needs Contingency: An AI Tutor That Escalates and Fades on What the Learner Does. arXiv preprint. https://arxiv.org/abs/2609.22993 doi:10.48550/arxiv.2609.22993

q3 · i? · causal · r2

Within-subjects comparison (N=131, 393 sessions) of three tutors on three Python tasks, analysed with linear mixed models and Holm correction. The contingent tutor exceeded the answer-on-request tutor on "effort spent on the approach rather than the typing"; against the question-only tutor approach effort was equivalent within 0.25 scale points (pTOST = .004).

"Tutor C exceeded Tutor A on effort (+0.46,𝑆𝐸 0.09,𝑝Holm <. 001) and on effort spent on the approach rather than the typing (+0.33, 𝑆𝐸 0.09,𝑝Holm <. 001)."

Hou 2026 (2)¶

Hou, X., Weng, Y., Yeh, C. H., Lee, D. L., Zheng, L., Li, F., Wang, W., & Liu, Y. (2026). Adaptive Scaffolding Needs Contingency: An AI Tutor That Escalates and Fades on What the Learner Does. arXiv preprint. https://arxiv.org/abs/2609.22993. doi:10.48550/arxiv.2609.22993

q3 · i? · causal · r?

Forced-choice comparative attributions collected once after all three tasks from the same 131 participants, tested against equal attribution with Holm correction. Tutor C was named most often for thinking hardest (52 of 131) and doing the most thinking (49), with Tutor B second and Tutor A third.

"Learners also named Tutor C most often as the tutor with which they thought hardest (52 of 131, 𝜒 2(2)= 16.8,𝑝Holm =. 002) and did the most thinking themselves (49,𝜒 2(2)= 13.2,𝑝Holm =. 011)"

Discussion¶

Related Claims¶

• In an immediate post-test design, participants with an unrestricted chatbot achieved significantly higher learning deltas than participants with a Socratic chatbot (d = 0.80) or an EEG-adaptive exercise system (d = 0.85) — the general claim this finding bears on, with the studies behind it

• Contingent scaffolding improves learning more than fixed or absent support. — the general claim this finding bears on, with the studies behind it

• A contingent tutor surrendered the full answer in one session in sixteen, against one in six for the question-only tutor — related
• Frustration under the contingent tutor sat between the answer-on-request assistant and the question-only tutor — related
• Contingent tutoring held delegation at the level of an answer-on-request assistant while delivering more artifacts, one decision at a time — related
• Perceived task load and germane load did not differ between the contingent tutor and either baseline — related
• On a text channel, fading holds when the learner's turn is aimed at the decision under support, regardless of its depth — related
• Scaffolding improves learning — a broader claim this one bears on
• In TBLT, teachers perceive scaffolding collapse, a shortcut paradox, invisible inequity, and task realism corrosion — related
• When AI assistance is withdrawn, student performance tends to decline, revealing a lack of internalized learning strategies — related
• Teachers most often perceived AI tutors and study buddies as academic support at a scale no single teacher can provide — related
• Low contingent support increased task effort when independent working time was short; with long independent working time task effort decreased under both contingencies but less so with high contingency — related