On a text channel, fading holds when the learner's turn is aimed at the decision under support, regardless of its depth¶

Claim · All claims
Evidence · 1 study · causal r? · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Turns not aimed at the decision under support were followed by concession 40.3% of the time against 28.8% for aimed turns; depth of the turn did not order the concession rates. → 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

q2 · i? · causal · r?

Content analysis of 515 coded fading episodes from Tutor C's 551 fading occasions, scored against later concession in the log. "turns not aimed at the decision under support" were conceded twice as often as aimed turns, and the gap held within rungs; depth did not order outcomes (a bare assertion at 20.4% was safest, a request to build at 33.7% worst).

"turns not aimed at the decision under support were followed by concession 40.3% of the time (72 open decisions) against 28.8% for aimed turns (1,084), a difference of 11.5 points, 95% bootstrap interval on participant [2.1,22.4]"

Discussion¶

Related Claims¶

• Students' application of teacher support predicts accurate answers in small-group work — related
• A contingent AI tutor places metacognitive demand equal to a question-only tutor and higher than an answer-on-request assistant — related
• The agent's praise was indiscriminate affirmation: validation followed about half of participant turns regardless of what the participant said — related