Recently, artificial intelligence (AI) has transformed university teaching and research by providing advanced tools that optimise data analysis, automate repetitive tasks, and improve the accuracy of decision-making by students. However, the implementation of AI in academia faces several challenges and limitations, including the bias inherent in the algorithms it employs, the need for interpretability in automated decision-making, and the reliability of the associated data. These factors and limitations could hinder the incorporation of AI tools into both (1) students' study procedures and (2) the assessment of the knowledge they acquire. In relation to these two issues, the main objective of this study has been to increase students' awareness of these difficulties so that they can optimise the incorporation of AI into their study procedures.

The conceptual framework of this research consisted of ‘exploring the opportunities that generative AI (ChatGPT) can offer in the acquisition and assessment of knowledge in various subjects in the agroforestry and environmental engineering degrees taught at the School of Forestry Engineering and the School of Agricultural Engineering at the Polytechnic University of Madrid.’ To this end, a methodology based on student work in the classroom was developed, consisting of two tasks. (Task 1) Design and completion of a survey to determine the current use of AI at the university. A group of students designed this survey through several face-to-face sessions.

Subsequently, these students encouraged a broader ‘target’ group of UPM students to respond to the survey (via the Google platform). (Task 2) Several groups of students from various courses duplicated the responses in a set of assessment tests following the procedure below. First, they completed each test based on their knowledge and by consulting the bibliography, and then they repeated each assessment test using the responses generated by AI (ChatGPT).

Preliminary results obtained in Task 1 indicate that 73 students, primarily studying science and engineering (63%), responded to the survey on the Google platform. The majority of these students use the AI tool regularly (54% of students), mainly ChatGPT (62% of the time). These students primarily use AI in their academic activities (92% of queries made to AI), both in their daily studies (66% of AI use) and when completing their assignments (70% of AI use). This population of students has learned to use AI either through self-teaching (32% of users) or through informal training (27% of users). The primary perception among the surveyed students is that AI will have a positive impact on their work (48% of respondents), although most are aware that its use can lead to errors (62% of users). As a result, most respondents (48% of users) verify the answers they obtain from AI.

Preliminary results obtained in Task 2 have indicated that the grades obtained by students using their own knowledge and consulting the bibliography have been more variable than those obtained using ChatGPT (with a standard deviation 60% higher when not using AI). However, the average grade obtained using ChatGPT was 12% lower than the average grade obtained by these students when they did not use AI.

In conclusion, this paper has shown the importance of students (and teachers) being aware of both current trends in the use of AI at university and the limitations of these tools when used without supervision of their results.

Keywords: Artificial Intelligence, forest engineering, agricultural engineering, environmental engineering, test-based assessment.