HCD.26:11 - SoTA-Echoing

Edwards et al., 2025 supplies a bounded design line connecting awareness, externalization and process prompting. The study involved 52 pre-service teachers in 14 groups; trained humans simulated proactive voice intervention, while a separate reactive chat used GPT-4. It found no significant increase in lexical alignment following prompts, and its revised prompt design remained a proposal.

Sections 4.1–4.4 adapt the useful construction while requiring actual uptake and consequence. Compared with an automatic reflection prompt, the intervention is selected from the work; compared with unassisted continuation, it can supply a missing occasion or question. Coordination burden and task preparation from the collaborative cognitive-load account constrain that choice. Reconsider the trigger when it is ignored, disrupts useful action or addresses the wrong cause. Neither cited account validates the constructed cases or establishes an autonomous agent’s equivalent performance.