Use of AI in exams with limited credits

TokenAI

Marion GUILLARD Mathis DUCHER Kathleen SOMERS INSA LYON, département Télécommunications

Training / education / pedagogy

We would like to study the **permission** of an AI tool in exams in the same way as for calculators. The target audience would be students only. Last year, we encountered a similar situation where we were authorised to use documents and AI during an exam. In this simulation, students who do not use AI during the exam will get a bonus. Those who do use it would have limited access to it, like a rare resource in the form of a credit limit.

This situation is interesting because students are always faced with exams. In the AI-permitted exams, the professor noticed that the students who obtained the highest marks were the ones who didn’t use them at all. But we think that forbidding AI in an exam might not be the solution. After all, we will be allowed to use it in our professional lives. Therefore, we would like to encourage students to enhance their critical thinking skills by writing accurate prompts within a credit limit (e.g. token, carbon footprint, etc). This could be implemented as part of a course to teach students how to use AI strategically. However, it might cause tensions among the students (those who are at ease with AI and those who aren’t), within the teaching community (pro-AI vs anti-AI) and the IT management in terms of responsibility. Concerning the impact, AI could clearly lead to dependence during exams if it is used for each one. Besides, it might not help students to deepen their understanding; worse, it could lead to superficial learning of the concepts.

As future telecommunications engineers, we want to take part in this challenge because we would like to contribute to the development of an ethical approach to AI. During our training, we use AI every day for coding, solving exercises and correcting errors. However, we are not fully aware of the limitations of using this tool. We want to learn more about environmental issues and find eco-friendly ways to use AI. We also want to learn how to weigh up the pros and cons by reading materials and reflecting on them. Finally, we are eager to see what suggestions the other teams have come up with.

The first solution we suggest is to allow the use of AI during the exam, but with certain restrictions. Students will be given a certain number of tokens, and the consumption of these will be determined by the number of questions, not the length of the questions. The AI will only be used as a chatbot, and it will not be possible to upload any files or pictures. Students who do not use AI at all will receive a bonus. On the Moodle platform, for example, there is a timer during online MCQs that indicates when the exam is ending. Here, a pop-up message will appear to indicate how many tokens are left. Once the professor has corrected the exam, we will analyse the relationship between the number of credits used and the grade obtained by the student. Furthermore, we will analyse the number of prompts and how frequently they are used. The aim is to encourage students to fine-tune their prompts. We hope that this solution will raise awareness of the fact that AI is a limited resource that needs to be used wisely. Finally, after submitting their exam, students will receive their carbon footprint for their interactions with the AI.