On AI and assessment
On 30 November 2022, the company OpenAI launched ChatGPT in an openly accessibly version for a limited period. Neither this tool nor other similar AI tools are new, but it appears to have been in connection with this launch that awareness of the technology and its potential truly hit home within higher education globally.
The tool’s ability to generate well-written, longer texts containing (at times) entirely correct answers to the questions posed could have huge consequences for how we assess our students. These pages aim to provide concrete tips and advice for how to prevent and detect AI-related attempts at deception during examinations.
They therefore have a consciously narrow perspective on AI, which is a much broader and multifaceted field. It is important to remember that AI may prove in time to be a valuable teaching tool for both teachers and students. A broader approach to AI, teaching and assessment, where the focus will shift to teaching and learning and to how AI can be used productively in teaching, will be pursued later in 2023.
What will you find on these pages?
On the page About ChatGPT and similar tools, we briefly and simply describe the basic principles behind how these systems work, what limitations they currently have and how they are likely to develop. Some of the risks of AI in connection with assessment are also outlined here.
The page Prevent prohibited use of AI offers tips for ways of working which strengthen the academic learning environment over the long term and which help render all types of cheating less relevant – including how teachers and students can use AI in a teaching context.
The page Hinder and detect prohibited use of AI offers a number of concrete proposals for working methods that make it less attractive to attempt to cheat in connection with an examination.
The page More information about AI and assessment contains links to more in-depth material and articles that could be useful in this context.
Finally:
The technology has made great strides, but it is still possible to determine in various ways whether a submitted piece of work has been produced independently or is the result of prohibited aids. This will without doubt require adjustments and changes to be made to examination practice, and there are also bigger challenges for certain subjects and courses than for others. If the University and its teachers take action now, there is time to develop sustainable ways of working and forms of assessment that will allow students to be assessed in a secure way even in the future and that offer effective and improved support to their learning.
If the current debates helps us to strengthen academic culture, review components for assessment and how we assess them, and if resources are indeed reallocated to more in-depth feedback, then it may turn out that AI, quite unexpectedly, truly contributes to a marked and qualitative improvement in higher education, even before we have seriously explored what the tools themselves can offer in different subject areas.
About the text
The text has been written, at the Deputy Vice-chancellor's request, by Mats Cullhed at the Unit for Academic Teaching and Learning, mats.cullhed@uu.se. A reference group has contributed with some text, suggestions, comments and corrections. University members of the reference group have been Beáta Megyesi, Christian Sköld, Cilla Häggkvist, David J.T. Sumpter, Emma Lundkvist, Heléne Andersson, Sonja Bjelobaba, Thomas Nygren, Ulrika Svalfors och Åsa Kettis. Anton Sánchez Sulejmani, chair of Uppsala Student Union, has served as student representative. English translation by Luke Halls.