Revision after chatbot feedback predominantly results in successful uptake, indicating effective error correction¶

Claim · All claims
Evidence · 1 study · associational r2 · q2
Currency · newest study 2026 · checked against its sources 2026-10-19 (automated, 1 of 1 entry) · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? When learners revised after feedback, the next state was successful_uptake with probability 0.69, and successful_uptake (7.25% of events) significantly exceeded unsuccessful (1.66%) or no uptake (1.55%). → Steve Woollaston 2026

Evidence¶

Steve Woollaston 2026¶

Steve Woollaston, Brendan Flanagan, Yuko Toyokawa and Hiroaki Ogata. (2026). Penny: Transition Network Analysis of Learner-Chatbot Interactions in Scaffolded EFL Writing. Preprint, accepted at the 1st Workshop on Transition Network Analysis (TNA) at LAK '26. https://arxiv.org/abs/2607.14575 doi:10.48550/arxiv.2607.14575

q2 · i? · associational · r2

Coded behavioural analysis of 63,328 events from the same 119-learner log dataset. Learner revisions following feedback were classified as successful, unsuccessful, or no uptake, with LLM classification validated by researchers (Fleiss' κ of 0.71); "successful_uptake (7.25%) was observed significantly more often".

"successful_uptake (7.25%) was observed significantly more often than unsuccessful_uptake (1.66%) or no_uptake (1.55%), suggesting that learners frequently and effectively incorporated the feedback provided."

Discussion¶

Related Claims¶

• Enacted Feedback yields higher workflow-specific uptake of AI-generated feedback than Directed and Self-Directed Feedback workflows — related
• Enacted Feedback produces more post-workflow revisions per resource than the comparison workflows — related
• In the Enacted Feedback workflow most resources moved from AI feedback to suggestion selection, while optional AI assistance was rarely entered — related
• Proactive multimodal feedback significantly increases feedback uptake compared to control — related
• Learner-chatbot EFL writing interaction follows two dominant behavioural loops: a Revision Loop and a Chat Loop — related
• Qualitatively, AI-peer integrated feedback was associated with a broader repertoire of revision operations and more agentive cognitive strategies than peer feedback alone — related
• EFL proficiency significantly shapes chatbot interaction: high-proficiency learners over-engage in dialogue while low-proficiency learners over-rely on feedback requests — related
• Customized AI agents generate feedback with higher accuracy and specificity than general-purpose agents in self-directed instructional design tasks — related
• A design-based study with 42 undergraduates reported a 15% improvement in rubric alignment and faster correction cycles from proactive agentic feedback loops — related
• High-acceptance groups made more effective use of AI feedback during revision, resulting in higher accuracy scores — related
• Classified examples are more often correct than generated examples during learning, but feedback does not improve classification — related
• Selecting a suggested option in the scaffold was associated with higher compliance for four out of twenty scaffold options in the generalized condition. — related