Assessing knowledge in a world with generative AI
On these pages, you can read more about how to minimize the potential impact of generative AI on your assessment. Begin with deciding an approach, then what actions to take. Actions can roughly be divided into two categories: Preventive measures or hindering actions. These actions are not mutually exclusive, and you should actively consider both perspectives in your teaching.
Background
Generative AI technology has a potential impact on assessments in higher education and most likely will it affect the way we assess the students' work from now on. One reason is that longer texts on specific topics now can be generated quickly using generative AI applications online. Other assessment materials can also be generated, for example program code. Cheating is therefore more accessible and often cheaper than before, but using a third party to produce assignments is not a new problem in itself. The difference is that with generative AI tools online, everyone can now access it, and it is free to use.
Decide your approach to generative AI
What approach to generative AI should you implement in your course: total prohibition, free usage without supervision, or somewhere in between? All have advantages and disadvantages, so it depends on the learning activity. The approach should be aligned with the learning outcomes and the type of knowledge or abilities taught and assessed.
Decide your approach to generative AI
Prevent and discourage prohibited use of generative AI
Preventive measures have the purpose of discouraging students from using generative AI in a dishonest way, mostly with clear communication and by putting less pressure on the students. However, the most important preventive measure is to create a culture among students at KTH where cheating is not even considered an option.
Prevent and discourage prohibited use of generative AI
Hinder and detect prohibited use of generative AI
By adapting the implementation of an exam, you can hinder the use of generative AI and increase the likelihood of detection. Design questions that generative AI have a hard time answering and test them with an AI. Restrict the exam setting with proctored exam or complement the exam or project with oral presentations. You can also combine these actions with detection tools, although such tools are flawed. All approaches require you to set clear boundaries on what is prohibited use of generative AI.
Hinder and detect prohibited use of generative AI