Students stop developing their own analytical capacity
Generative AI offers significant learning improvements but enables easier plagiarism and reduces student critical thinking capacity through over-reliance on technology. Students increasingly use AI for writing and brainstorming while professors struggle to assess authentic learning; social isolation from reduced peer interaction compounds concerns.
- Institute of Public Policy report by Steffen Hoernig and Paulo Trigo Pereira, released September 2024
- Students increasingly use generative AI for writing and brainstorming; professors struggle to detect plagiarism and assess authentic learning
- Identified risks include reduced critical thinking, social isolation, inherited bias in training data, and data privacy violations
- Recommended solutions: field-specific institutional guidelines, investment in fair and transparent AI systems, stakeholder involvement, and promotion of digital literacy and ethical awareness
Portuguese research highlights generative AI's educational benefits while warning of plagiarism, reduced critical thinking, and social isolation risks. Experts say proactive institutional guidelines and ethical awareness can mitigate these challenges.
A research team at the Institute of Public Policy in Lisbon released a report this week that captures the central tension of generative AI in universities: the technology is genuinely useful, but it comes with real costs that institutions are not yet equipped to manage.
The report, authored by professors Steffen Hoernig and Paulo Trigo Pereira, acknowledges that generative AI has transformed what's possible in higher education. Students use it to refine their writing and brainstorm ideas. Teachers can design courses more efficiently, assess student work faster, and communicate with larger cohorts. The learning outcomes themselves can improve when AI is used thoughtfully—it can walk students through concepts step by step, push them to think deeper, expose them to explanations they might not have found on their own.
But the same technology that enables all this also enables something else: students can now submit work that is not their own, and it is harder than ever to tell. Because generative AI produces personalized responses based on each user's specific commands, plagiarism detection becomes a game of diminishing returns. A professor grading fifty essays faces a new kind of exhaustion. The report identifies this as one of several interconnected risks that universities must now confront.
The deeper concern, though, is what happens to the student's mind. When a student can offload the hard thinking to a machine, the muscle of critical analysis atrophies. The report notes that generative AI can either strengthen critical thinking—by presenting problems that demand real engagement—or it can erode it, by making it too easy to accept whatever the algorithm produces as sufficient. Students begin to trust the machine more than their own judgment. They spend hours interacting with screens instead of sitting across from classmates or professors, and that social isolation carries its own cost. The ability to work in groups, to negotiate ideas with peers, to develop social competence—these things are harder to practice when you are alone with a chatbot.
There is also the problem of inherited bias. Generative AI learns from databases available online, and those databases are not neutral. They contain gaps and distortions. Women's research is underrepresented in fields like science, technology, engineering, and mathematics, and that gap is baked into the training data. When a student relies on AI to summarize a field, they are learning a version of that field that is already skewed. The technology can amplify injustice without anyone intending it to.
The report does not counsel universities to ban generative AI or to pretend these risks do not exist. Instead, it argues that institutions must move first, before crisis forces their hand. Universities should develop guidelines for AI use—not one-size-fits-all rules, but frameworks tailored to each discipline, because a literature seminar and an engineering lab face different questions. They should invest in systems that are fair, transparent, and secure. They should bring students, faculty, and technical experts into the room when designing these policies, not impose them from above. And they should cultivate what the authors call a culture of ethical awareness, so that using AI becomes a conscious choice rather than a default.
The authors are clear: these risks are not inevitable. They can be prevented and minimized if universities acknowledge they exist and decide to act. The question now is whether institutions will move fast enough to shape how this technology integrates into their communities, or whether they will spend the next five years reacting to problems they could have seen coming.
Citas Notables
Generative AI offers enormous potential for students, teachers, and institutions, but the associated risks cannot be ignored.— Steffen Hoernig and Paulo Trigo Pereira, Institute of Public Policy
These risks can be prevented and minimized if higher education institutions are aware of their existence and willing to confront them.— Steffen Hoernig and Paulo Trigo Pereira, Institute of Public Policy
La Conversación del Hearth Otra perspectiva de la historia
Why does the report focus so much on plagiarism when the deeper issue seems to be about thinking itself?
Because plagiarism is the visible crisis that will force universities to act. It's the thing that will show up in a dean's office first. But you're right—plagiarism is a symptom. The real disease is that students stop developing their own analytical capacity. Plagiarism is just the moment when that becomes undeniable.
The report mentions that AI can actually improve critical thinking if used the right way. So the technology itself isn't the problem?
No, the technology is neutral in that sense. The problem is the default. When a student has the option to think hard or ask the machine, most will ask the machine. Universities have to make the hard thinking the path of least resistance, or it won't happen.
What does it mean that students are spending more time with machines than with each other?
It means they're not learning how to disagree with a person, how to sit with discomfort in a conversation, how to build something together. Those skills don't transfer from human-machine interaction. They're social, and they require friction.
The report mentions bias in the training data—women's research being underrepresented. Can that actually be fixed?
Only if universities are intentional about it. They'd have to audit what their AI systems are recommending, push back on incomplete results, teach students to question what the algorithm shows them. It's possible, but it requires work that most institutions aren't doing yet.
So what would a university that gets this right actually look like?
It would have clear rules about when AI is a tool and when it's a crutch. It would involve faculty in designing those rules, not impose them. It would invest in training people to use AI well. And it would measure success not by how much AI they've adopted, but by whether their students are thinking better.