Considerations for studying with generative AI
The use of generative AI in graded assignments and exams should always be based on the course-specific information, but when studying, it is your own judgment that matters. Here you will find general guidelines to help you meet legal requirements and ethically use generative AI. The guidelines are adapted for students at KTH based on guidelines from DIGG (the Swedish Agency for Digital Government) and the EU Commission.
Always start with course-specific information
First and foremost, you must follow your courses' course-specific information about how generative AI may be used in graded assignments and on the exam. No information on this page takes precedence over what your teacher says applies.
Tip! Adapt your use of generative AI to how you may use it in graded assignments and exams, so you are properly prepared.
Read more about Course-specific information on generative AI .
General guidelines
You are recommended to use these guidelines as a basis for your use of generative AI as they will help you use generative AI ethically and legally. The guidelines are based on DIGG's guidelines for generative AI in public administration but adapted to be more relevant for students at KTH. We have also included some information from the EU Commission's ethical guidelines for the use of AI in education. You can find more information and links to this after the guidelines, under the heading "Further reading".
The guidelines are as follows:
- Only use generative AI when it is relevant and appropriate.
- Have a human-centric and respectful approach to generative AI.
- Work critically and with human review (“human-in-the-loop").
- Take GDPR and copyright into account.
- Be transparent about how you use generative AI.
More details on how to meet these guidelines can be found under the following headings.
1. Only use generative AI when relevant and appropriate
Before you start using generative AI, you should ask yourself the most basic question ("Question zero"): what are you really trying to accomplish? And as a follow-up question: is generative AI the best way to achieve what you're trying to accomplish?
Choose from other tools
Only use generative AI when generative AI is needed. There are many other tools that can be more suitable and energy-efficient for specific tasks.
For example:
- Search engines.
- Existing templates.
- Spell checker.
- Image banks.
Weigh desired effects against potential undesirables
Consider the potentially undesirable effects that may result from using generative and weigh them against the positive, desired effects.
Examples of possible undesirable effects:
- Reduced collaboration and exchange with other students as generative AI is used as a sounding board instead.
- That your unique way of expressing yourself is washed out of your texts.
2. Have a human-centric and respectful approach to generative AI
You should take a human-centric approach to generative AI, that is, you use generative AI so that you respect the identity, privacy, and dignity of people.
Focus on retaining and developing skills
Generative AI can perform tasks that you otherwise would need to develop skills to complete, which may put you at a disadvantage in the future. Vary your use of generative AI to give yourself opportunities to practice the skills that generative AI performs for you.
Examples of skills:
- To develop your own voice in writing by writing without AI.
- Setting up and solving problems on your own.
- Picking out what is important enough to include in a summary.
3. Work critically and with human review (“human-in-the-loop")
If you use a generative AI tool, you should use it as an assistant that supports you in your work, not as an expert on any topic. The tool may be missing important details or creating false information (called "hallucinating"). You are the one responsible for all decisions and assessments of what is appropriate and correct, and you also have empathy and social skills that the tool lacks.
For example:
- Include diversity and different points of view in your prompt (instruction).
- Always end with human review and correction of the result. For example, correct:
- misinformation (hallucinations)
- prejudice
- distortion of facts.
Tip! Feel free to edit the AI-generated result to put your spin on it, for example to get your voice into a text.
4. Take GDPR and copyright into account
GDPR and copyright are two laws that directly affect what you are allowed to include in your prompt to a generative AI tool. The following two lists explain how to most easily take GDPR and copyright into account when using generative AI.
GDPR:
- Remove all personal data from the material you provide to the generative AI tool. What is actually meant by personal data? (imy.se) .
- Make sure that you have set up so that your material is not used for training the tool. For ChatGPT, for example, you can find how to do it in their FAQ: How do I stop my chats from training ChatGPT? (help.openai.com) .
Copyright:
- Material created by generative AI is not copyrighted.
- Avoid creating material that may infringe copyright. For example, by avoiding creating images in a specific artist's style.
- Exclude copyrighted material from your prompts or get permission from the copyright owner before using it.
If you are interested in more details, you can find them on the intranet: Legal aspects of generative AI (KTH Intranet) .
5. Be transparent about how you use generative AI
Tell us when you use generative AI in your own material, preferably also how and why. If a course provides more detailed instructions on how to report your use of generative AI, make sure to follow it.
Further reading
We have adapted existing legal and ethical guidelines to better suit students at KTH. We have used: