AI Changes Exams at Universities: Students Must Prove What They Know

After realizing that oral warnings from professors are insufficient and that students skillfully find ways to secretly use artificial intelligence in their academic work, many global educational institutions have decided to take matters (or, more accurately, return them) into their own hands and clearly define how and for what students may, and may not, use AI during classes and exams, as well as explain the consequences of violating these rules.

For example, for the upcoming academic year, the Columbia Law School in New York has explicitly prohibited students from using artificial intelligence in exams, when drafting legal arguments and conducting legal analysis, as well as in writing texts, although they can still use its tools as assistance in learning and understanding, researching, and correcting grammatical errors. In their explanation, Columbia University emphasizes that they are aware of the rapid development of technology, but that through their rules among law students, they aim to encourage critical thinking and independent judgment.

Similarly, some European universities have already taken this path, but such official documents are still rare at Croatian universities, and the use of artificial intelligence is mainly regulated by recommendations and guidelines from professors, varying from faculty to faculty. However, one of the few Croatian educational institutions that has translated recommendations for using AI into explicit rules for students is the Zagreb School of Economics and Management, which has adopted the Regulation on the Use of Artificial Intelligence in Teaching and Student Work.

Early Adopters

As explained by the Vice Dean of ZŠEM’s undergraduate studies, Goran Oblaković, their goal is to promote responsible use of artificial intelligence in initial research, analyses, and idea exchanges.

– Of course, any presentation of a text written with the help of artificial intelligence as one’s own, student work is academic fraud, or plagiarism. AI must not replace original thought, critical analysis, or academic research. Every use of AI should be transparent and cited in accordance with academic citation guidelines – notes Oblaković.

Interestingly, ZŠEM has noticed that there are fewer below-average seminar papers from students, but at the same time, the percentage of excellent papers has not increased. This shows that AI can help to some extent, but, as Oblaković points out, ‘it cannot replace original thought and critical thinking’.

– All courses at ZŠEM have always had an oral component for assessment, but since the emergence of AI, we have intensified oral examinations – he adds.

Algebra Bernays University has also adapted its knowledge verification methods due to artificial intelligence, as they believe that due to the development of this powerful technology, it is even more important to clearly define learning outcomes. There, however, no regulation has been adopted, and the use of artificial intelligence tools is not completely prohibited for students. However, as they explain, the student must be the author of the work they submit and must be able to explain, argue, and defend its content.

– Our approach is not aimed at a complete ban on the use of AI tools, but at clearly distinguishing their permitted use from situations where artificial intelligence replaces the independent academic work of the student. AI tools can be used as assistance, but not for writing the work or parts of the work that the student then presents as a result of their own thinking. The question of source reliability is also particularly important. An AI tool is not a scientific source, and the student remains responsible for verifying all facts, references, and conclusions included in their work – they explain from Algebra Bernays.

In this regard, exams now require students not only to provide a correct answer in content but also to show how they arrived at that answer and why they believe a particular solution is justified.

– A specific example of adaptation is that we have limited the possibility of using AI tools in part of the exam, even when the student is allowed to use certain official documentation or other predefined sources. This way, we want to make a clear distinction between the ability to find, understand, and apply relevant professional sources and leaving the task-solving process to a generative AI system – they add.

In doing so, they do not seek to ‘outsmart’ artificial intelligence in knowledge verification, as they say this would quickly become unsustainable.

Support in Thinking, Not Replacement

Both institutions emphasize that students are increasingly and better using AI tools, as the tools themselves are becoming more sophisticated. For example, they can help them learn and explain concepts they do not understand, as well as structure learning materials. However, critical thinking has never been more important. For instance, Oblaković highlights that many studies, including their own at ZŠEM, show that students make poorer decisions with the help of artificial intelligence, even as their self-confidence grows.

– Artificial intelligence confirms their initial ideas, and they do not strive to critically consider other possibilities. AI is a very useful tool, but proper use alongside critical thinking is crucial – explains Oblaković.

From Algebra Bernays, they also note that problems arise when artificial intelligence begins to be used as a replacement for thinking, and they observe similar behavioral patterns.

– The student can quickly receive a convincingly formulated answer, but without a real understanding of the content. That is why we consider it important to develop awareness that AI tools can support thinking, but not replace it. Additional education is needed in verifying the credibility of information, critically evaluating the results generated by AI, and proper referencing and transparent indication of the use of AI tools. This is not a topic relevant only to student work but for the entire academic community – they believe.

Although many universities and faculties still do not have strict regulations, this does not mean that there are no consequences for inadequate use of artificial intelligence. Moreover, the same rules for plagiarism apply to misuse because in both cases the work is no longer original.

– If the use of AI tools results in the submitted work no longer being the result of the student’s independent work or if the true authorship of the content is concealed, such behavior is viewed within the framework of existing academic integrity rules and can have the same consequences as other forms of unauthorized assistance or inauthentic academic work. Thus, we are not creating a special ‘AI offense’, but we apply existing principles of academic integrity to new technology – they explain at Algebra Bernays.

AI and MBA

Given that the quality of education today determines the labor market of tomorrow, some world universities have gone so far as to introduce artificial intelligence into their MBA programs. For instance, Wharton, the business school of the University of Pennsylvania, is launching a special MBA track and undergraduate specialization in artificial intelligence in business starting this academic year. Domestic universities and schools currently do not offer anything similar, but of course, their goal is to prepare students for a business world where the use of AI tools is almost as taken for granted as working in Excel once was. At ZŠEM, they are monitoring technological changes and preparing students for the labor market and new demands of the corporate sector in certain courses, while at the level of departments, programs, and institutions, they continuously align changes.

– We believe that our lecturers, thanks to their rich work experience, can best assess what needs to be adapted at the course level – notes Oblaković.

Algebra Bernays University has introduced artificial intelligence into all its study programs, specifically in those subjects and areas where students are expected to know how to use these tools, in collaboration with employers and the needs of the real sector.

– The goal is not just to teach students to use specific tools, as they change very quickly and it is difficult to predict what they will look like in a year, five, or ten years. In the long term, it is more important to develop the ability to recognize problems, define tasks well, assess the results obtained, and understand the consequences of applying a particular solution. Regardless of which solution the AI tool proposes, the responsibility for evaluating that solution, making decisions, and how it is applied remains with the person using it, whether they are a manager, expert, or employee – they conclude at Algebra Bernays.