E It has long since become undeniable: Generative artificial intelligence (AI) has pulled the rug out from under political science examination practice. Our field urgently needs reform. We propose combining digitalization with humanization in such a way that AI promotes human learning and at the same time human abilities are tested fairly again.
Large Language Models can produce acceptable texts – and thus deliver the end product at the push of a button, through the creation of which people actually practice the skills that are considered the learning objectives of a political science degree: developing and analyzing facts, recording and structuring technical debates and applying them to concrete objects, and writing down results in a coherent, understandable and precise manner.
Written work – term papers, later bachelor’s and master’s theses – is still considered the most important form of examination. We use this to evaluate how well the students achieve learning goals, aggregate this into a grade and give the students feedback to promote their learning process.
Students must learn to use AI productively and critically at a university level
AI is changing this situation: Some students continue to write their own work. Others, however, submit generated texts. Lecturers can hardly tell the difference – there are no watermarks or reliable technical verification methods. A fair assessment is hardly possible anymore. Students, on the other hand, face general suspicion. The relationship of trust is disturbed.
So far only disparate solutions
The problem has been identified (although not yet in its entirety from all of them). This is shown by internal university recommendations for dealing with AI, specialist debates and conversations with students and colleagues. But attempts at solutions remain disparate. They range from a total ban on AI to exam tasks that require and evaluate the use of AI.
In between there are attempts to make use more difficult or easier to discover. Sometimes people resort to AI-safe – traditional – formats such as handwritten exams and oral exams. With their inconsistent and inadequate response, universities are not doing justice to the changing era. The situation is intolerable. So what to do?
If we continue to want students to learn what they are supposed to learn – namely reading and understanding academic texts, sorting and rearranging information, thinking and writing – and if we continue to have the right to fairly evaluate their own performance, then in our view there is only one way out: the radical redesign of university learning and examinations, accompanied by intensified support and a reduction in individual examinations.
Our reform proposal combines the consistent and thoughtful digitalization of universities with the simultaneous humanization of studies. Consistent and reflected digitalization means that students must learn to use AI (and other digital tools) both productively and critically at a university level. Most students already use AIbut very few are sufficiently informed about its genesis, function, potential and limits. Humanization means that university assessments must be designed in such a way that only the abilities of the people being examined are actually assessed.
What to do?
For universities, this means, firstly, that they have to fundamentally revise their study and examination regulations. The focus must be on oral examinations and written work that the students can demonstrably write themselves. Secondly, for the latter, the necessary infrastructure must be built. PC pools are needed in which students can write texts for their assessments without AI. This ranges from term papers that are no longer written at home to bachelor’s and master’s theses. This is more time-consuming and requires more support than current examination practice.
Thirdly, it is important to reduce the number of examinations. Not every seminar has to end with a twenty-page paper. Fourthly, it is important to ensure that universities have the human resources necessary for the AI era. The supervision ratio in many political science institutes is already too bad for the optimal type of learning, which is based on individual approach, feedback and revisions – i.e. accompaniment of the process instead of just the evaluation of the end product.
If students even in the AI era If you still want to acquire the necessary skills to pass exams at the end – with texts you have written yourself – the support of the learning process must be much more closely monitored. Fifthly, there needs to be an accompanying systematic exchange at the university level so that successful approaches can spread quickly.
The use of AI has long been as present in research and teaching as it is in the professional fields for which political science studies are ultimately intended to prepare people. It is therefore urgently necessary to convey to students their targeted use as a tutor and, overall, a powerful tool for scientific research and work.
At the same time, the acquisition of one’s own cognitive skills during studies would – finally again – be the focus of our reform proposal. Because the critical and productive use of AI requires a crucial element: human judgment. And this can only be developed by those who have previously acquired the relevant skills themselves.