Chat GPT - Opportunity or Danger?

UBTaktuell: What opportunities does ChatGPT offer?

Prof. Dr. Martin Huber: Important opportunities lie in the competence development of students. They can use the technology for better results, especially in research. Microsoft has now linked its search engine Bing with the latest ChatGPT version 4. This software can also help to develop new research questions in dialogue with the software, for example, and to match them with existing research databases in the future. In addition, students are sensitised to working with artificial intelligence and practise dealing critically with computer-generated texts. However, they need to be guided for this. This requires a rethinking of teaching concepts and new competencies among teachers.

What does this mean for teaching?

ChatGPT will change teaching permanently, as all subjects will have to redefine what their core competencies are that they want to develop in students. Central teaching questions will be: What do students need to know and understand, how can they analyse problems of the subject and what do they need to develop in a creative and scientific process? To what extent can AI support me in this - for example in research, text revision and analysis? What subject-specific competences do students need in order to be able to critically analyse and evaluate the AI-generated information? Because the goals of teaching are changing, we are also designing the teaching and learning processes for students accordingly, and adapting the forms of examination accordingly. Currently, the Centre for Higher Education Teaching (ZHL) is offering and developing appropriate training formats for all teachers. There are also plans to integrate students into the discussion process.

What are the concrete plans?

We will integrate subject-specific reflection on the impact of AI in all degree programmes. In addition, the University of Bayreuth is also developing its own study programmes that deal with the application of AI in science, the effects of AI related to humans, for example, with regard to ethical questions and the emergence of knowledge. There are currently over 100 researchers at the University of Bayreuth whose research focuses on informatics processes and artificial intelligence.

What risks do you see in text-generating search engines?

I don't want to call it a risk. But of course we have to keep an eye on the consequences: Some areas of written, scientific work could be replaced by current search results from ChatGPT, because currently the software generates passable - but in my opinion not very elegant - answers. The software will get even better in the future and can access current data from the internet. Students could use the software to generate parts of a written paper with it.

Are we talking about plagiarism here?

It is already the case that text passages generated by software must be marked as such. Plagiarism detection software already exists for this purpose. The faculties are currently discussing the forms of examination that will be required in the future, which will react to the use of ChatGPT. Like teaching, ChatGPT will also change the logic of our examinations, because we want to test students' academic competences in the examinations. To this end, we will increasingly use individualised forms of examination and examination concepts that require students to transfer their knowledge and apply it situationally. We will also expand the exams to include critical reflection on the use of artificial intelligence in the research, work and writing process.

So just reproducing what you have learned is a thing of the past?

That depends a lot on the subject, of course. But yes, I think we need to require our students to ask the right questions rather than just give the right answers. New competences in dealing with automated writing aids are needed for all those involved - teachers and students - so that at the end of the course we also release AI-competent graduates in all subject areas.

That ChatGPT changes the way we examine is obvious. What about the way we teach? Does that have to change too? In what way? What about the content of teaching?

Let's let AI answer that for once. If we ask this question to bing, the search result in chat mode looks like this:

"The way of teaching in universities should change with ChatGPT so that artificial intelligence is seen as a help rather than a threat. Some possible changes are:

- Teachers should encourage students to use ChatGPT creatively and critically to improve, extend or vary their own texts, but not to replace or plagiarise them.

- Teachers should enable students to distinguish ChatGPT texts from human texts by teaching them how to check the quality, sources and reasoning of the texts.

- Teachers should encourage students to use ChatGPT as a learning partner who answers their questions, provides explanations or conducts dialogues, but not as a teacher who prescribes or evaluates their content.

- Teachers should ask students to use ChatGPT ethically and responsibly by making them aware of the legal, moral and social consequences of their actions and by showing them the limitations and risks of artificial intelligence.

ChatGPT can thus be an opportunity to enrich, diversify and individualise teaching and learning in universities if used properly."

I want to comment on and classify this Bing chat result from 15.2.2023: The software clearly presents which possible changes are conceivable. In doing so, the software focuses on the teachers' perspective of what they should do vis-à-vis the students. It should stimulate, enable, encourage and call on them to use it. Unfortunately, the software does not address the risks that arise when students are less comfortable with AI use and could be left behind as a result. Likewise, the dimension of data protection is not mentioned, because students must not be forced to disclose personal data in the form of user accounts or the like. The chat result also does not address the limits of the software, because currently it can - to our knowledge - primarily only access free, digital educational information and not information behind paywalls or non-digitised knowledge stocks. However, it appreciates the ethical handling of the results, because it is important to know that ChatGPT was fed with texts from the internet, whose origin and e.g. political agenda is not always directly recognisable. Consequently, we have to reflect critically on all the results. It is to be welcomed that with ChatGPT4 the cross-references to other sources are listed and we get access to the supposed background texts. It is interesting to note the last sentence which addresses that there are opportunities for individualisation for students. Further thought should be given here to what this might look like in concrete terms.