Can artificial intelligence be used to write essays? How can AI support learning? These are the questions currently under consideration, as ChatGPT, an AI application known as a large language model, has gained an increasingly broad user base since its release in November.
In connection with the swiftly growing popularity of ChatGPT and other similar user interfaces, the University of Helsinki has published guidelines pertaining to the use of AI in teaching and studying. The guidelines cover principles of use, cheating and equality.
“A lot of technical solutions that boost studying have become available, but the scale of the current paradigm shift is such that it must be taken into consideration in teaching as well as its planning and organisation. Indeed, I would like to see a broad discussion in degree programmes on its possible effects on the content and forms of education,” says Jaakko Kurhila, the University’s Chief Digitalisation Officer.
Eyes on the horizon
At the University of Helsinki, artificial intelligence is seen as an opportunity that provides both teachers and students with means to develop in their work and studies. Kurhila is pleased that those involved in university education have been quick to familiarise themselves with the possibilities of text-generating AI.
“Trials in teaching situations at this stage depend largely on the teacher’s personal interest, since the ChatGPT application was only released a few months ago. For example, many of the fellows of the University of Helsinki’s Teachers’ Academy wish to be at the forefront of development. People in all fields of science should be alert to the change,” says Kurhila.
A support for thinking, not a substitute
For the uninitiated, it may seem that, from now on, students just have to provide a handful of keywords and a frame for an AI solution, after which they will soon receive an essay or thesis ready for printing and submission.
“This is not the case. Artificial intelligence chooses statistically sensible words on the basis of the content used in its training, which is why it easily generates ostensibly convincing nonsense. And AI of course has no responsibility for the text it produces. The person who utilises AI is responsible for accuracy,” Kurhila points out.
Professor of Computer Science Hannu Toivonen emphasises that large language models should be used in support of your own thinking and brainstorming. In contrast, using AI as a substitute for thought does not get you far.
“I personally wouldn’t use artificial intelligence to produce text. After all, it’s not just about the need to verify facts, but also about hidden assumptions – the unspoken and possibly even entirely inaccurate assumptions in the text generated by language models. It’s much easier to write the text yourself from scratch than to read through something that may look fine, but may contain problems that are difficult to perceive and understand.”
The utilisation of large language models is already changing teaching arrangements at a fairly rapid pace. For example, students could previously complete maturity tests in computer science in their own time. With the new University guidelines, they will now be completed at invigilated examination sessions.
“As teachers, we have to consider assignments that do not allow the misuse of language models – or situations that require flexibility,” Toivonen notes.
AI literacy to accompany media literacy
The opportunity to make use of artificial intelligence in studying has so taken educational institutions by storm that there has been little time in just a few months to educate either teachers or students in the matter. According to Hannu Toivonen, members of the University community should acquire AI literacy, which includes an understanding of the principles of AI as well as its limitations, opportunities and threats.
“Previously, we have talked about media literacy. Now it is time to learn to take a critical approach to language models and the texts generated by them, while also learning to recognise their benefits.”
At the Teachers’ Academy, a working group focused on AI in teaching is currently being established. The University’s policies and guidelines will be updated as experiences accumulate.