Verify to learn: Inspectable AI use in introductory mechanics for engineering students

Authors:
Publisher website
Google Scholar
Abstract
This study reports a short classroom intervention designed to make students’ use of artificial intelligence (AI) visible while promoting critical verification in introductory mechanics. The activity was implemented in a first-year, calculus-based physics course for engineering students with two parallel sections ( n = 42 and n = 33 ). Using a crossover structure across the work–energy and impulse–momentum units, students completed the activity under AI-supported and non-AI classroom configurations. An additional AI-supported implementation was carried out in rotational dynamics. Each one-hour session included conceptual pre-task and post-task items, an isomorphic problem-solving activity, and a structured verification section. For the conceptual component, no immediate changes were detected in the work–energy and impulse–momentum units. For the problem-solving component, the analysis compared students’ initial attempt with the final revised solution submitted after verification. This final solution is interpreted as the quality of the product produced within the activity, rather than as an independent post-test of problem-solving ability. In both crossover units, final revised solution accuracy was higher in the AI-supported configuration than in the non-AI configuration, with the largest initial-to-final changes appearing in more demanding multi-step items. Written critical comments were common in AI-supported sessions, although they are interpreted as evidence of documented engagement with verification rather than as a direct measure of evaluative judgment. The intervention offers a practical way to incorporate AI into ordinary assessment while shifting attention from answer retrieval to reasoning, checking, and evaluation.

This publication has 23 references indexed in Scilit: