Abstract:

Proper use of AI tools requires the application of critical thinking skills, such as analyzing information, evaluating evidence, and making well-reasoned decisions, to varying degrees, depending on the task, the reliability of the AI output, and users’ individual learning habits. At times, these skills are applied extensively (e.g., checking sources or refining prompts), while in other situations they may be used only minimally (e.g., accepting the first answer without verification). This study explores the development of critical thinking skills among students in the Master’s Degree in Food Science and Technology (MSc-FST) and the Bachelor’s Degrees in Biotechnology (BSc-BIO) and Agri-Food and Rural Engineering (BSc-AGR) at the Universitat Politècnica de València by examining their patterns of AI tool usage.

Information was collected through a seven-items questionnaire which was created using Google Forms and administered online during an on-site session. Participation was voluntary and anonymous, and informed consent was collected from all participants.

Results suggest that a considerable proportion of students demonstrate behaviors associated with critical thinking, such as verification, reformulation, and cross-checking of information. Specifically, 80% of MSc-FST students, 67.8% of BSc-BIO students, and 46.2% of BSc-AGR students reported that they always or almost always revise AI-generated answers. Likewise, 90% of MSc-FST students, 71.4% of BSc-BIO students, and 53.9% of BSc-AGR students frequently reformulate prompts, indicating iterative questioning, a key component of critical thinking. These behaviors align with students’ perceptions of AI accuracy: none of the MSc-FST students and only 32.2% of BSc-BIO and 7.7% of BSc-AGR students considered the first answer provided by Copilot or ChatGPT to be accurate or to require little adjustment, reflecting a degree of skepticism and a tendency to question AI outputs.

Cross-verification, one of the strongest indicators of critical thinking, was also widely practiced. None of the MSc-FST students and only 4 of 56 BSc-BIO students and 2 of 13 BSc-AGR students stated that they do not compare AI-generated answers with other sources such as textbooks or class notes. This high level of validation reflects a healthy, critical approach to AI-based information. Consistently, the main criterion students used to judge the acceptability of AI outputs was their agreement with class notes, showing that they relied on trusted academic materials.

Students also expressed positive views regarding the usefulness of AI tools: all MSc-FST students (100%), 89.3% of BSc-BIO students, and 53.9% of BSc-AGR students rated AI assistants as quite useful or very useful for academic purposes. Moreover, between 50% and 89.3% of students across the three programs believed that AI tools significantly improved the quality of their academic tasks.

Overall, the results indicate that most students, particularly those in the Master’s Degree program, exhibit moderate to strong critical thinking behaviors when interacting with AI tools. They check facts, question initial answers, and compare information with trusted academic sources. This suggests that AI tools are not only helpful for learning but can also support and strengthen critical thinking in higher education.

Keywords:

Critical thinking, AI-assisted learning, trust in AI outcomes, AI-student interaction.