Skip to main content
To KTH's start page

How to talk to your students about generative AI

Do you wonder how to best communicate your approach to generative AI to your students? We recommend you inform them of the general rules early, then develop more detailed class expectations in collaboration with your students. During the course, encourage continuous discussions, be transparent with your usage and model the behaviour you expect from your students. For more details and real-life examples, keep reading.

Inform early

Early on in your course you should give your students the general rules on how and when to use generative AI in different course activities. These rules should preferably be available before the course starts, so you can include them in the course memo.

Examples of course memos at KTH

In the course-PM for “DD1320/DD1326 Tillämpad datalogi” there is an addition to the Code of Conduct specifically for generative AI. It reads (translated from Swedish):

“Code written by AI is counted as code written by one of your friends in your stead – you are not allowed to present it as your own work in a laboration.

Feel free to use AI to ask for explanations and examples.

In some laborations you will be encouraged to try out ChatGPT, but it is never mandatory.”

In the course-PM for "DD1390 Programsammanhållande kurs i datateknik” the following information can be found under the heading “Ethical approach” (translated from Swedish):

“Generative AI tools are powerful tools that, used correctly, can help you learn more effectively. Don't use AI if it reduces your own learning. Check in the assignment instructions if and how you may use generative AI. Do not trust AI-generated text to be correct. If you have used generative AI in a submission, you must always disclose it.

In the reflections of this course, you may not directly use text generated with AI tools. However, it is permitted to use AI tools to get ideas for your reflections.”

Develop class expectations in collaboration with students

To expand on your general rules of using generative AI, you should have your students discuss the benefits and limitations of AI in relation to the course’s learning objectives. To quote Arnold Pears in episode 48 of the podcast “Fikasnack om Framtidens utbildning” (translated):

“If you used an AI tool to undermine your achievement of a learning objective in a course, then you have used AI in the wrong way. If you use the same AI tool to enhance your understanding and to achieve the learning objective to a greater extent than you could do alone, then you have used AI in a positive and good way.”

This discussion is a golden opportunity to gather students' input on class norms or guidelines for AI use, providing them some agency and ownership in course policies. Encouraging students to take responsibility in their own learning in this way may help them feel more invested in meeting your expectations for their coursework. This is also one of the principles to reduce cheating at KTH, which you can read more about in the Updated report! - Promoting learning and preventing cheating .

Real-life example

In the Lunch ‘n’ Learn “AI and examination”, Jane Bottomley explains how she explored and influenced how students use writing tools, such as generative AI. She includes the reflection task she used and some takeaways from it.

Lunch 'n' Learn: AI and examination

Encourage continuous discussions

If your students use generative AI during your course, they will gain practical experience in what works and what does not. To help turn this into usable knowledge, you should facilitate authentic dialogue between students about the benefits and limitations of AI. For example, encourage them to start an assignment by considering the advantages, costs, and ethical questions of using generative AI in said assignment. You can also encourage this during class discussions and in post-reflection activities.

These continuous discussions combat over-reliance on AI tools and helps students avoid common problems. For example, AI generated text can contain sentences students do not fully understand, yet keep, as they fear editing them will worsen the text. These texts are often more impersonal as well. To quote one of Jane Bottomley’s students in the “AI and examination” LnL:

Student: “I don’t like how these tools rewrite my text, it does not sound like me.”

Be transparent and model expected behaviour

It is important to be open with your students about your knowledge (even if it’s limited). You cannot be expected to be an expert in generative AI technologies, nor is it useful to pretend to be one. In many cases, you and your students will be exploring the educational possibilities of generative AI together and there might be a need to revisit previously established guidelines.

Your students will be watching how you approach and use generative AI, which gives you an opportunity to model expected behaviour. If students see that you are being inquisitive, careful, and ethical in your approach to AI use, they are more likely to engage in a similar manner. For example, disclosing if and how you have used generative AI in the course material. You can even vocalise your thoughts while you demonstrate AI-related skills to explain your choices. For example, when writing a prompt to input into an AI tool you can consider what details to add out loud and ponder the quality of output it generates.