FAQ: AI in education
In order for each teacher to be able to assess the potential of generative AIs, particularly those specialising in text production, it seems important that they should be able to test these tools and form their own ideas about their possible uses in their teaching (e.g. help in creating teaching materials, support for students, integration of these technologies into the tasks required of students, etc.) or simply about the consequences of their use by students in their courses (e.g. interview with a chatbot for questions of comprehension related to their course, production of summaries and other revision materials, etc.). ) or simply on the consequences of their use by students as part of their course (e.g. interview with a chatbot for questions of understanding in connection with their course, production of summaries and other revision materials or even help with writing academic papers).
For those who would like to test a large-language model (LLM), we can suggest, for example:
- Try asking it to produce a text that you need to respond to (for example, an email, a memo or a proposal).
- To ask him/her questions related to your teaching, that a·e étudiant·e would be likely to ask in the context of understanding one of your courses.
- To ask him a few questions from your previous exams and to assess the quality of the answers obtained.
You will notice that the more information you give the AI, the better its result. It doesn't know anything about you, but it can extrapolate, analyse and link information remarkably well if you give it something to work with.
In view of the accessibility of these tools, their current performance and expected progress, and given that it is difficult to determine what they produce, a ban seems neither relevant nor realistic.
Therefore, it is important that teachers and students are trained and informed about the characteristics of these tools, their risks and their use;ristic of these tools, on their appropriate use as well as on the issues of academic interest they entail and the institutional rules on the matter.
Generative AI tools can be good allies for teaching if they are used with full knowledge of their advantages and limitations. Their use requires a certain degree of expertise and an ever-evolving critical mind. These tools also require the utmost caution and vigilance with regard to the way in which personal data is processed and used (see section on data protection and confidentiality).
Users of this type of technology should also be aware that the quality of the results depends very much on the quality of the prompts (instructions) used; of the prompts (instructions) used and that it is therefore important, if one wishes to make use of these technologies, to master "The art of the prompt".
The Centre de Soutien à l’Enseignement offers a watch on the uses of AI génératives in higher education, which is available here: Pedagogical corpus: generative AI and higher education.
For teachers
For teachers, and in a non-exhaustive manner, generative AIs can provide benefits in the following aspects:
Enrichment of course syllabi
Generative AI tools can be used as partners in the creation or enrichment of teachers course syllabuses. They can, for example, help to draw up or improve course objectives (learning outcomes) or suggest a variety of teaching activities or even evaluation proposals. By way of example, Professor Mitchell Weiss of Harvard Business School presented some simple ways of using AI to create or refine a course syllabus. Nevertheless, he reminds us that we should never rely solely on AI to do this work. In fact, it is of course the teacher who must be behind the creation of the courses and who must carefully check the results of the AI.
https://hbsp.harvard.edu/inspiring-minds/if-your-syllabus-needs-a-refresh-generative-ai-can-help
Design of course materials
Text management tools can, for example, be used to produce summary sheets on subjects related to a course, glossaries built around specific terms linked to lessons or suggest situations that illustrate concepts covered in courses. An example can be found on this page.
Paid solutions such as H5P or NOLEJ offer automatic creation of interactive books from one or more sources of information (handouts, videos, articles, etc.). These solutions are currently being tested by the Teaching Support Centre ; feel free to contact Jean-François Van de Poël if you'd like to try them out.
- Example of production with NOLEJ: https://live.nolej.io/preview/a43b9b37-9406-40a7-8f52-298c9512dab6
- Example of production with H5P: https://universityoflausanne.h5p.com/content/1292115087769586327
In thess cases, the quality of the resource depends on the quality of the provided sources.
Image management tools such as Midjourney, DALL·E or Adobe Firefly can be an interesting source of illustration for course materials.
Designing exam questions or formative assessment questions
Generative AI tools can suggest exam questions or questions linked to formative assessments offered to students.
Wooclap and Wooflash, two tools available in the portfolio of digital tools intended for teaching at UNIL, currently offer a feature that allows students to manage their own learning; This allows users to create MCQs and flashcards from downloaded documents, text or even videos:
Access to these two tools requires your Switch edu-ID credentials.
Designing complex academic assignments
Generative AIs, if used well, can become good allies in the creation of complex academic work. For example, they can be used to help manage case studies, suggest standard solutions or provide correction bars.
As an example, we suggest you take a look at the discussion by Professor Thomas Steiner, from the HES-SO: Rethinking your exam in the light of chatbots.
For students
For students, chatbots powered by LLMs such as ChatGPT offer the possibility of having a learning companion available at all times to answer questions relating to the courses they are taking. But it is imperative that they are made aware of the limits of these models and of the importance of maintaining a sharp critical mind regarding the answers and information obtained.
The use of this type of technology should not be a substitute but rather an aid.
Students can, for example, use generative AI to:
- Initiate academic work in connection with the production of a text, an analysis or computer code, for example (within the limits of what their teacher·e prescribes in terms of the use or citation of génératives AIs).
- Illustrate their work
- Rendering personal feedback material (flashcards, MCQs, summary sheets, etc.)
- Propose improvements to the form of their work
- Simulate a dialogue, an argument (in a foreign language for example)
It is important that the uses they implement are in perfect compliance with the rules in force at UNIL on matters of intellectual property and citation, as well as the rules governing the use of the material; (see Directive 0.3).
At present, and despite numerous publications and announcements, the quality of the data still leaves much to be desired. The fact that they are not sufficiently reliable means that they cannot be used as a basis for restrictive policies or the application of sanctions.
First of all, only academic work carried out outside the walls of the institution represents a potential source of concern for teachers (e.g. dissertations, term papers or programming). In fact, examinations and academic work carried out under supervision do not give students access to a learning tool that can be used to improve their knowledge and skills; a générative AI tool if this is not authorised, with the exception of those produced on computer on the "Open Book" format.
Validity problems may therefore arise above all in the case of examinations or assignments organised without supervision. However, this mainly concerns work requiring simple restitution, a synthesis of content, a summary of works or the production of computer code. On the other hand, work that requires critical analysis, personal reflection or argumentative justification is currently not very accessible to AI.
Most generative AI tools require the creation of an individual account, which a student who does not wish to create one cannot be forced to do (see Which AI tool does UNIL make available to its community?).