<![CDATA[Physics problem solving depends on how solvers represent situations, select principles, apply equations, and evaluate results. This comparative-descriptive mixed-methods study examined responses to five heat-and-temperature problems from 30 Grade XI students at an Islamic boarding school, three physics experts, and one generative AI model (three runs per problem; 180 solution units). Solutions were analyzed using a five-dimensional rubric, separate strategy/checking codes, and think-aloud evidence from six students and three experts. Equation-first strategies appeared in 70 of 150 student solutions (46.7%). Fourteen of 15 AI outputs (93.3%) followed a structured-mechanistic sequence; nine of 15 expert solutions (60.0%) were explicitly coded as principle-first with checking, while six were concise but coherent. The largest expert–student difference was in Physics Approach (1.82 points), and the smallest in Mathematical Procedures (1.28 points). The multistage task yielded the lowest mean total for all groups (students 15.00, AI 21.33, experts 22.33; maximum 25), consistent with demands for process decomposition. Because the analysis concerns observable outputs, similar written structures do not establish equivalent reasoning processes. Instruction should require principle justification before equations and checks of units and plausibility, with AI solutions critically compared against expert solutions.]]>
Copyrights © 2026
