Source-informed choices and limits

The best-known line used here is task- and interpretation-sensitive evaluation, with examples that distinguish rival rules and observations that preserve assistance and continuation. HCD.18 and HCD.19 carry the comparisons, worked consequences, and source-specific reopens. These are professional adaptations; the studies do not validate this profile or its local ordinal anchors.

Wesenberg et al. (2025) supply failure evidence for ambiguous worked examples in bounded topics. McCrudden et al. (2022) and 2024 inform the required relation between principles, examples, and self-explanation. The resulting choice is to make the relation consequential in a task; it is not a universal ordering rule.

Noetel et al. (2022) inform mechanism-specific representation repair. Sinha and Kapur (2021) support distinguishing preparation followed by resolving instruction from unsupported difficulty. Wisniewski et al. (2020) support testing feedback’s function rather than its presence. None supplies a universal local threshold, fixed reader count, or whole-book adequacy finding.

Bastani et al. (2025) inform the separation of assisted and later independent performance in their studied arrangement. Krieglstein et al. (2025) discipline the timing and scope of experienced-burden observations. Bavaresco et al. (2025) supply task-dependent limits on substituting LLM judgements for human ones. Their practical consequence is qualified evidence, not simulated human learning or a ban on useful support.