How mining engineering students checked ChatGPT answers

Updated Nov 11, 2026

Using ChatGPT can mean accepting an answer, checking a calculation, challenging an assumption or trying to recover a forgotten concept. A study published in Higher Education on 10 November 2026 traces those differences through graduate mining engineering students' accounts and interactions. For teams reviewing digital learning, it offers a useful question to ask after whether students use AI: what do they do with its answers?

Follow the task as well as the opinion

Rana Alhaj-Bedar, Nicole Cunningham, Zach Agioutantis, Steven Schafrik, Ali Moradi and Pedram Roghanchi examined master's and doctoral students in a US mining engineering department. Participants completed tasks involving a conveyor calculation, an open-pit mining explanation and an underground-mining decision.

The researchers brought screen recordings together with written reflections, interviews and surveys. This let them compare students' explanations of their approach with the interactions recorded during each task. Our digital-learning pathway connects this question with the wider experience around a tool: its purpose, the activity and the support students need.

Distinguish reliance from verification

The qualitative analysis identified five interaction patterns. Some students relied on the answer with little checking. Others relied on ChatGPT while actively verifying its output, worked with it as a collaborator, used it selectively with scepticism, or used it to develop their understanding.

Those distinctions describe different actions. Checking a calculation against existing knowledge is different from asking the system to provide more detail. Challenging whether the supplied information supports a decision is different again. Students described directing prompts, assessing answers and deciding when their own knowledge should take priority.

Keep changing approaches visible

Participants' approaches varied across tasks. The researchers traced shifts between reliance, verification, scepticism and collaboration within individual students' work. A single description of someone as an AI user would lose that movement between activities.

The underground-mining task also exposed a specific difficulty. It included misleading or invalid information; ChatGPT proceeded as though the question were valid, and most participants did not identify the task's mistake. Checking an answer therefore also involves examining the information and assumptions in the question.

Read what follow-up questions are doing

Students' follow-up prompts served different purposes: requesting an explanation, checking the reasoning, seeking missing factors or developing a concept. The researchers interpreted interaction time and question counts alongside recordings, reflections and interviews. The content of the exchange helped explain what the activity meant.

For a local evaluation, ask students to describe a recent task, what they asked the tool to do and how they decided the response was acceptable. Invite an example of something they checked, changed or rejected. Our account of engagement measures and student dialogue offers a related question about connecting recorded activity with students' explanations.

Use the distinction when reading comments

For Student Experience and digital education teams, our practical suggestion is to retain the task and action in the comment. A response about saving time may describe drafting, calculating or checking; a response about confidence may concern understanding the subject or trusting an answer. Read those accounts separately before deciding what support to review.

Take the selected issue to students and the relevant teaching team. Ask whether the next step concerns task instructions, opportunities to practise verification or the route to academic help. The analysing-comments pathway helps keep the original question and the interpretation connected.

FAQ

What did the study examine?

Graduate mining engineering students' perceptions and interactions with ChatGPT during calculation, explanation and decision tasks.

What was useful about combining the evidence?

The researchers could read students' reflections and interview explanations alongside their recorded interactions, distinguishing several ways of relying on and checking AI answers.

What could an institutional team ask students next?

Ask for a recent task, the role the tool played and how the student decided which parts of its answer to use, check or reject.

References

Alhaj-Bedar, R., Cunningham, N., Agioutantis, Z., Schafrik, S., Moradi, A. and Roghanchi, P. (2026). Cognitive engagement and students' perceptions of ChatGPT during discipline-specific problem solving in mining engineering graduate education. Higher Education. Published online 10 November 2026. https://doi.org/10.1007/s10734-026-01806-y.