HCD.11:11 - SoTA-Echoing

Working question: how can an assessment support a claim about a person’s contribution when tasks, assistance and later use differ?

Selected line — adopt and adapt. The 2014 Testing Standards, Chapter 1 tie evidence to an intended interpretation and use. Expanded evidence-centred design connects tasks and observations to inferences and makes learning during assessment relevant. HCD.11 adapts that line into the short chain in §4.2 and the separate attempts in §5.1.

Serious alternative: score the final work product with one rubric. That is cheaper when the product’s origin is settled and the inference concerns only that product. Where help or a non-compensable criterion changes the human claim, the same rubric can give a confident wrong answer. Keep its useful criterion-based scoring, but reject the substitution of total score for the intended contribution.

The AI tutoring experiment by Bastani and colleagues adds a bounded counterexample: assisted practice gains in high-school mathematics did not guarantee later unassisted performance. It supports keeping support conditions visible, not an adult-workplace effect estimate or a ban on AI.

The selected line changes §4.3’s observation of help and §4.4’s separate judgements, at the cost of only the discriminating evidence needed for the use. These sources guide design; they do not validate this local assessment. Reopen the task or inference when local evidence shows that it rewards a proxy, misses a material error or attributes a tool’s contribution to the person.