A contingent tutor surrendered the full answer in one session in sixteen, against one in six for the question-only tutor¶

Claim · All claims
Evidence · 1 study · causal r2 · q3
Currency · newest study 2026 · checked against its sources 2026-10-17 (automated, 1 of 1 entry) · not reviewed by a person

Subclaims¶

q3 i? The welfare floor fired far less often under the contingent tutor than under the question-only tutor, which could give content only through that floor (on a learner's surrender, or after a set number of support rounds) and which the authors describe as releasing on surrender. → Hou 2026

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

Session end-state measure from the log (whether the whole answer was given away), stricter than artifact delivery. Tutor A produced the complete program in three sessions in five; the floor fired under Tutor B more than twice as often as under Tutor C, and distress followed the same order.

"Tutor B refused, and released only when the learner gave up, which one session in six reached. Tutor C put the request on the ladder, asking about the decisions the build turned on, and the asks ran out in one session in sixteen"

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
• Productive cognitive offloading to AI can shift into cognitive surrender, abdicating learner intellectual agency — related
• Contingent tutoring held delegation at the level of an answer-on-request assistant while delivering more artifacts, one decision at a time — related
• A contingent AI tutor places metacognitive demand equal to a question-only tutor and higher than an answer-on-request assistant — related
• Frustration under the contingent tutor sat between the answer-on-request assistant and the question-only tutor — related
• Contingent scaffolding improves learning more than fixed or absent support. — a broader claim this one bears on
• Perceived task load and germane load did not differ between the contingent tutor and either baseline — related
• The leakage manipulation fires: ConvTutor leaks far more than PedTutor on the primary base, while probe accuracy saturates and cannot distinguish conditions — related