Questions and answers on GAI tools
On this page, you will find an assortment of questions asked during webinars about GAI tools. The questions have been answered by members of Lund University's working group for generative AI tools.
Questions about function
Generative AI tools differ from other AI tools because they can generate new text, images, and audio in response to simple questions written in natural language. They have "learned" patterns and structures from existing examples of text, images, or audio and then generate new versions.
GPT-3 does not, but other tools do – such as Perplexity and Bing AI.
Yes, if you have a substantial amount of written material and provide the tool with a structure. This will become common as AI tools become more integrated into everyday software we use, such as search engines and Office programs like Word or PowerPoint. Focus on what's important here: Is it about having written all the words, or is it about collecting and analyzing data? Many discussions will be needed on this in disciplinary meetings and with publishers and research funders.
There is no doubt that AI tools will become the primary way we find and summarize information. We will want to use these tools more and more, but the information they provide can be unreliable and biased. Therefore, we will need to identify and apply ways to review and verify the results.
Questions about teaching and learning
The opportunities vary across different disciplines. Some things you can do with the help of these tools include generating ideas, improving texts, analyzing data, and automating tasks - for example, suggesting text for routine emails.
It's a good idea to test the tool yourself and find out the basics so that you can decide how to work with these tools in your teaching and discuss them with your students.
We can incorporate these tools into teaching in many ways. Here are some examples:
Digital literacy: Ask students to use the tools to answer a question and discuss the results in small groups. Ask them to identify key points to look for to verify the results.
Idea generation: Have students use AI tools to brainstorm ideas on a difficult question, and then discuss how suitable and feasible the ideas are.
Structuring: Ask students to use AI tools to generate a structure for a written assignment, such as an essay. Then, have them identify what additional information they need to start writing. You can also ask them to write a few paragraphs together.
Questions about examination and assessment
No. Some have observed that the output can be repetitive and dull, but human-written text can also be. There is no reliable way to recognize the output, and the pace of development is such that even if we provided tips now, they would become outdated very quickly.
There is software that claims to detect the use of various GAI tools. These are being tested both here and at other universities. However, we cannot currently recommend them as they have not been verified to comply with data protection regulations. You should not upload student work to any systems that have not been approved by the Legal Division at Lund University.
There is no simple answer to this question. Developing new approaches requires a lot of time and thought. Here are three examples of things you can do:
- Emphasize the value of students' own work - highlight what a university education and individual thinking contribute to students' personal and academic development.
- Encourage criticism of the use of GAI tools as well as discussion and peer review of the students' progress in the work.
- Contact the educational unit your faculty collaborates with to see what workshops and courses are available to support you.
If you decide to allow the use of GAI tools, you can ask students to explain what they have used and how, for example, "I used ChatGPT to generate computer code for running the analysis presented here." If this replaces the work you expect students to do on their own, you can develop the assignment. For instance, you can ask them to explain how they verified the quality and accuracy of the output from the GAI tool, or what understanding they gained (or lost) from using the tool. You can also complement a written assignment with a brief oral examination to ensure that students have achieved the learning objectives.
Questions about rules and guidelines
At Lund University, it is permissible to use generative AI tools in education if you believe they can contribute to or facilitate learning. However, there are some important things to consider. As a teacher, you need to:
- be able to confirm that students have achieved the program's and course's learning objectives
- be aware that sharing students' work with GAI tools is not allowed
- ensure that every student has equal access to the GAI tools you recommend
- inform students about the risks of sharing personal information or copyrighted material with the tools
- inform students about the rules for using GAI tools in the course or program.
If you do not want students to use GAI tools, you must clearly state this in the exam information and grading criteria.
It is a disciplinary offense if:
- the tool was an unauthorized aid according to the course syllabus or assignment instructions
- the student pretends that they performed the work themselves.
Check the university's regulations and the Higher Education Ordinance for more details, but remember that it is difficult, if not impossible, to detect the use of these tools.
Yes. Librarians have already started organizing this - for example, the University of Queensland already has several citation formats for GAI sources.
Probably not right now, although the definitions of "authorship" may change over time. GAI tools can be considered a source. If a significant portion of the work includes AI-generated text, this should be acknowledged, and you can let people decide whether you should take authorship ownership.
Do you have questions about GAI-tools?
Email Rachel Forsyth: digital [at] education [dot] lu [dot] se (digital[at]education[dot]lu[dot]se)