Examination and Generative AI?
Generative AI and examination
Generative AI's (GenAI) great strength lies in finding patterns and creating new text, images, program code, etc. based on given frameworks and patterns, such as summarizing, analyzing, or comparing texts of various kinds. The tool can also produce material that, at first glance, may be perceived as both convincing and correct, but which, on closer examination, turns out to contain both pure factual errors and overly generalizing assumptions. At the same time, GenAI is a fantastic resource when it comes to helping a text writer to both get started with their text production and support the linguistic quality, not least for those who are bilingual and do not have the language of the examining text as their mother tongue.
In other words, used correctly, the tool can support and help our students in their work, but also create challenges when it comes to distinguishing the individual student's knowledge during examinations. It is therefore absolutely necessary to write out what is allowed and in what way GenAI may be used in the course and its examination elements, which is completely in line with FUN's updated guidelines for course memos and examinations.
When you are going to write down what applies to your particular course and its examinations, there are three different starting points regarding its use.
Specify requirements for use
If you want to require students to use GenAI, it is important that you ensure equal access for the students and that it is not associated with costs that were not known before the start of the course or that students are forced to use services that can jeopardize their privacy. At the time of writing (March 2024), all students have access to Copilot from Microsoft via their student account, which can be a good starting point in the choice of tool if you intend to set requirements for the use.
Setting requirements for the use of the study can be a conceivable alternative if you focus on learning outcomes that focus on evaluating, theorizing, predicting and reflecting.
The advantage of setting requirements for the use of GenAI is that you can measure the student's subject-specific knowledge in relation to the use of GenAI. The disadvantage is that you must ensure that you create a legally secure examination, i.e. that all students have equal opportunities to carry out the examination in question. An additional challenge is that these types of tools have different capabilities depending on whether you pay for them or use free versions. Finally, you will most likely need to rewrite your course objectives for this to be possible.
Allow Usage
If your goal is to measure knowledge in the form of the student analysing, explaining and comparing information of different kinds in their examination, it may be wise to allow the use. If you go down this route, it is important that you also explain how the tool can be used and how to interpret and manage its results.
The major benefit of allowing the use is that you encourage students to research and take advantage of the power of GenAI linked to the subject area. The challenge for you as a teacher is, in addition to the equivalence aspect, that you need to reformulate your course objectives so that what you intend to measure is consistent with what is now possible to do. Also, don't forget that you need to carefully describe how students should report their use and thus maintain academic honesty.
Prohibit the use of
A third approach is to ban its use altogether. In order for this to be possible, the examination needs to be carried out with limitations and some form of supervision, for example in the form of written examinations or oral presentations on site. This can be a useful strategy if you primarily intend for the student to demonstrate knowledge in the form of being able to reproduce, identify, describe and define.
Banning the use can be a way to go if you primarily focus on factual knowledge and that what is reported comes from the student's own memory. The disadvantage is that the examination needs to take place for a limited time as its implementation needs to be supervised and that you do not get the opportunity to measure knowledge in a more realistic situation where access to GenAI is a natural part.