Will Ai make a project for me? Like universities and students approach artificial intelligence

For several years, artificial intelligence has changed the world of students faster than the ministry and deans are able to keep up with new regulations. In theory at the Warsaw School of Economics or at the University of Warsaw, there are already precise rules regarding the use of generative artificial intelligence. In practice … Everything is too too far from a particular faculty, promoter, and often even – a specific student group.

– From the perspective of didactics in SGH, regulations are rather conservative. It is allowed, for example, to support AI tools at the editorial office or brainstorming, but writing all work by AI, even with this fact, is prohibited – explains Dr. Albert Tomaszewski, an expert of AI Lab of the Warsaw School of Economics.

Similarly at the University of Warsaw: – The most important thing is to use these tools transparently – emphasizes prof. Katarzyna Śledziewska, head of Delab UW. The university promotes responsible, conscious use of AI and formally implemented recommendations regarding “Genai” in didactics. Promoters are to agree with students the scope of the permissible use of AI, and its use is prohibited during written exams.

And what does reality look like? Your regulations, and your daily practices.

“Shadow AI”, or how artificial intelligence students really use

Students are, as usual, one step ahead of the system. Despite the official prohibitions and declarations, there is a phenomenon in the academic environment that lecturers refer to as “shadow ai” – the use of AI in secret, except for the official circulation.

– According to the regulation, the use of AI should be reported, but nobody reports it because he does not want to report – says Dr. Tomaszewski

Student groups on Facebook, Discord or Messenger are full of “how to generate a paper, so as not to arouse suspicion” or “which detector AI is easy to cheat”. The instructions are famous for how to “prescribe” the generated text in your own words or pass through another paraphrase tool so that the AI-detector is stupid. One of the popular advice we found on the web was: “Generate in ChatgPT, then paste into the deeppl and let the Polish concealer pass.”

– In my opinion, there are currently no effective ways to detect such texts. The algorithms I have encountered have a high level of errors in both directions. They often do not detect the text created by AI, and on the other hand they can mark the original text as generated. Today it cannot be successfully checked – comments the SGH expert.

At the same time, as research at the University of Warsaw show, students are aware of threats. Many are afraid of unconscious plagiarism or AI mistakes. – Among the main reasons not using AI, respondents mention: lack of trust in the method of data processing, fears of errors, hallucinations and manipulation of content, as well as the risk of unconscious plagiarism – says prof. Śledziewska.

Not everyone trusts artificial intelligence immeasurably – some students use AI not to cheat, but to organize notes faster, generate an outline or to right style.

– Diploma thesors should together with students agree the scope and manner of using AI and monitor progress, while maintaining the author’s full responsibility for the content of the work – explains prof. Śledziewska, describing how you use AI when writing diploma theses

University under fire. “Submission tornado” and “flat” conclusions

Lecturers have no illusions: “submission tornado” – an avalanche of quickly generated, to pain similar work, is a new standard. – With tasks in class, such as papers, group projects or essays, you can see thematic convergence – 30 students submit work around several the same threads, because the algorithm does not generate too original ideas. Once the works were more diverse – notes Dr. Tomaszewski.

The superficiality, lack of deep analysis, “flat” conclusions – these are the most common hallmarks of texts generated by AI.

– Students transfer such work to the promoter, according to the principle “and it is enough”. If the text is of poor quality and not very diverse, it was probably created using AI – he adds.

Anti -plagiarism systems with anti -AII plugs in practice are not very effective. – If the system detects something, it is difficult to prove it, because a student can always deny. I do not support bringing everything to detectors – says prof. Śledziewska. The old, human method turns out to be much more effective: direct conversation and checking whether the student really understands what he wrote about.

– Detectors are not very effective, and trust and cooperation much more important – adds an expert in the University of Warsaw.

In narrow specializations, even the best LLM (Large Language Model, a large language model) does not pretend to be an expert: – In the case of narrow specializations, it is very easy to capture abuse – without in -depth knowledge of the topic, the text AI is superficial – notes prof. Śledziewska

New University’s approach: more research, less “essays”

Experts agree: AI writing text by AI becomes so simple that the mere “passing of written work” ceases to make sense.

– Key research should be: interviews, surveys, field work – this AI will not do it – says Dr. Tomaszewski.

Diploma theses should require originality, in -depth analysis and innovative approach to the subject, not just an efficient editorial office. More and more promoters are giving up classic written work for projects, presentations and even oral exams, during which it is easier to verify the real knowledge of the student. – Departure from written work is the simplest thing you can do, although it’s a bit avoiding a problem. I partially gave up my written work, because reading texts generated by AI simply does not make sense – explains Dr. Tomaszewski.

The approach to assessing students is also changing. – If we treat the diploma thesis as a common service, the project implemented together with the student, then the use of AI does not bother – supports the learning process. In this approach, rating for “five” or “three” loses its meaning. The road, development, cooperation and mutual trust are more important – notes prof. Śledziewska.

Academic future. Own AI tools

Artificial intelligence will not disappear from the academic world. On the contrary – it will be increasingly integrated with science, didactics and everyday students.

According to experts, universities should develop their own AI tools, adapted to Polish realities and protecting academic sovereignty. – For me, the presence of Bigtechs in the education system is a greater threat. I think that we should take care of the university’s sovereignty – the University of Warsaw has the potential to develop its own tools and models. The ministry should support such initiatives, promote cooperation with Polish solutions to maintain independence – emphasizes prof. Śledziewska.

But for now the race continues. Students are still inventing new ways to “bypass the system”, and the lecturers learn to recognize where the creative support of AI ends, and the academic fake begins. One thing is certain: – We return to the basics: to discuss, develop critical thinking and personal development, and not just checking information or memory – says the head of Delab UW.

The discussion about AI in education can be compared to how we use the calculator today, which makes life easier or how we approach the chess computer. Dr. Tomaszewski uses this analogy: “In chess, the computer defeated a man over 20 years ago, but in tournaments we still play without support, because the essence of the game is independence.” In education, similarly – AI can support, but will not replace human effort.