HCD.6:11 - SoTA-Echoing

The working question is: How should a practitioner design practice tasks that elicit a target human contribution and support meaningful correction and variation without mistaking surface realism for representation?

The selected best-known line combines sufficiently whole complex tasks, expertise-sensitive worked support, and decision-bearing variation. The maintained 4C/ID model and its fourth-edition presentation supply the serious whole-task design candidate: learning tasks, support for non-routine reasoning, timely procedural information, and additional practice for recurrent parts. Paas and van Merriënboer’s 2020 review (COG-05) supports worked examples and integrated information for novices while warning that unchanged guidance can become redundant as task knowledge grows. HCD.6 adapts these contributions into §§4.2–4.3 and the separation of whole-task and focused-part practice. The amount of help and the use of a repeated or varied task follow the learning purpose and starting capability; the sources do not prescribe a changed condition for every correction.

The serious ordinary alternative is a topic exercise or surface-isomorphic scenario: present the rule, vary the nouns and numbers, and score the final answer. It is cheaper to produce, but it can preserve answer-bearing cues, omit the later-Work purpose and interfaces, and reward either calculation without judgement or generic caution. HCD.6 rejects surface change as representation and adds the decision-bearing variation, human/support allocation, domain criteria, and correction task in §§4.1–4.6.

Brunmair and Richter’s 2019 meta-analysis (COG-03) supplies limit evidence that interleaving varies by material and similarity; it does not prescribe one mixed task set. Czyż, Wójcik, and Solarská’s 2024 motor-learning review (COG-04) supports testing delayed transfer while its applied estimate and cross-domain reach remain weak. Bastani and colleagues’ 2025 school-mathematics trial (AI-04) supports separating assisted practice from a later unassisted attempt; it does not supply an adult engineering effect or a universal AI ban. These sources change the task contrasts and evidence-purpose boundary rather than selecting one universal schedule.

The selected line costs more domain and case-design effort than renaming exercises, but it preserves action interpretability, legitimate support, critical errors, and the opportunity to learn from a changed result. Reopen it when stronger sources alter the whole-task, worked-support, variation, AI-help, or transfer boundary, or when representative use shows a lower-effort task design that preserves the same decision-bearing relations and evidence value.