Creativity in Artificial Intelligence: students, education, and homogenization
Artificial Intelligence is already part of the daily study routine at top universities like Yale and Bard College, and it is deeply changing how students speak in class and how professors assess what they truly know. The original CNN report draws a very clear picture: seminars that used to be full of different opinions now sound like one big chorus in unison. The vocabulary is more polished, the arguments more refined, but originality is giving way to answers that all seem to come from the same place: a chatbot.
In small classes, the scene repeats itself. While the professor asks a question about the reading for the day, some students lower their eyes to their laptops and start typing into an AI chat, trying to generate, on the fly, a well-structured comment. The answer may sound like a sophisticated analysis, but in practice it tends to repeat the same frames and reasoning patterns. What should be a lively debate almost turns into a sequence of mini standardized essays that sound correct, yet are only loosely connected to each student’s own experience.
Recent research in cognitive science reinforces this perception. Studies cited in the report show that large language models have been homogenizing the way we write, argue, and even think, along three main dimensions: language, perspective, and reasoning strategies. And this is not just a technical detail. When everyone starts relying on the same models to formulate ideas, the diversity of thought that used to emerge naturally in the classroom begins to shrink.
How AI is making debates more alike
One of the most visible effects is the way students speak and write. Professors describe a pattern: different papers, from completely different people, start to share the same neutral tone, the same paragraph and argument structure, the same type of introduction and conclusion. That personal flavor of language, with quirky constructions, unusual references, and even the typical slips of someone who is still learning, starts to disappear.
According to the study cited by CNN, this is not happening by chance. The models are trained to predict the most likely next word, based on a massive volume of text. Since that material tends to privilege dominant languages, cultures, and worldviews, the answers end up reflecting a narrow and biased slice of what humanity produces. Researchers call this the reinforcement of WEIRD points of view – an acronym for Western, Educated, Industrialized, Rich, and Democratic.
In practice, when a student uploads the class PDF into a chatbot and asks for a summary, an analysis, or a ready-made comment to use in a seminar, they are not just getting help with form. They are also importing a particular way of looking at the topic. Over time, if everyone in the class does this, alternative perspectives – more local, more intuitive, more grounded in personal experience – start to lose space. The result is that feeling, as one Yale student put it, that now “everyone kind of sounds the same.”
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
When even reasoning gets standardized
It is not just the text that becomes generic. The way of reasoning also starts to follow very similar tracks. AI tools are increasingly trained to display step-by-step chains of logic, explaining how they “arrived” at the answer. This is great for clarity, but it may slowly suffocate other ways of thinking: more intuitive, more visual, more connected to specific cultural backgrounds.
Researchers cited in the article warn that if a group interacts repeatedly with these systems, collective creativity tends to flatten compared to the same group working without AI. Instead of exploring multiple routes to reach an idea, students start relying on the same strategies the model uses, reducing the variety of mental paths.
The concrete impact in the classroom
At the universities portrayed by CNN, this effect has already become a practical problem. In economics seminars, for example, students report that, right before class starts, almost everyone uploads the assigned readings into a chatbot and asks for summaries, argument maps, or possible answers to questions the professor usually asks. When it is time for the debate, technically no one is completely lost, but almost everyone builds their remarks on top of the same bundle of interpretations.
Professors describe a strange mix: the floor for participation has gone up – very few people say something totally off-base – but the creative ceiling has dropped. Weird ideas, unexpected connections, and comments that go against common sense show up far less often. One of the instructors interviewed describes this as the disappearance of “strange, eccentric, and original” thinking that used to emerge even from students who were not the strongest in formal terms, but had strong intuition and the courage to take risks.
This also affects how students are assessed. In many courses, homework is no longer a reliable reflection of what a student actually knows. The same impeccable essay that looks impressive on paper may, in reality, have been heavily built with AI. And since the models can answer pretty much any open-ended exam question, it becomes hard to tell when a student has simply become a good “prompt operator.”
The frustration of those who avoid AI
Not everyone dives headfirst into this heavy usage. Some students, like those quoted at Yale and Bard, avoid using AI precisely because they are afraid of losing their own voice. They describe a double discomfort: on one hand, concern about the environmental impact of intensive use of giant models; on the other, the feeling that their classroom experience suffers when classmates use chatbots to fill discussion time with generic takes.
One student describes preferring to admit that she did not understand a certain text, rather than pretending she did with a flawless comment generated by a model. For her, the key to learning is truly engaging with the reading, connecting the material to her own story, and building critical thinking. When that is replaced by a canned speech, the discussion stops being deep and turns into something bureaucratic that just goes through the motions.
How professors are responding
Faced with this scenario, instructors in different fields have started to redesign courses, assessments, and rules for AI usage. At Yale, for example, there are institutional guidelines that make it clear generative tools may or may not be used depending on the course, and that automated AI detectors are not reliable enough to be used as the basis for punishment.
A few strong trends are emerging:
- Fewer laptops in class: many professors have limited or even banned computer use in certain meetings, prioritizing paper readings, handwritten notes, and face-to-face discussions.
- In-person assessments: there is a growing emphasis on in-class exams, essays written on the spot, presentations, and oral assessments, where the professor can see in real time how the student formulates ideas.
- Reused problems: some logic, computability, and technical courses have turned problem sets into question banks for in-person exams. Students may use AI on homework, but during the exam they will have to solve similar variations without help.
- Feedback instead of grades for homework: to reduce the temptation to outsource everything, some instructors have lowered the weight of take-home assignments, using them more as a basis for feedback and guidance than as decisive grades.
Some have also adopted individual oral exams, a practice a few professors were already using before the generative AI wave. In these meetings, students must explain solutions, defend their reasoning, and answer unexpected questions. It is a labor-intensive format, but one that makes it much harder to hide the true level of understanding behind a manufactured text.
From banning to guided use
Not everyone is choosing the path of outright bans. Some researchers advocate for a more sophisticated use of AI, positioning the model as a collaborator rather than a substitute for human thinking. Instead of asking the tool to write entire essays, for example, they use it to:
- test the consistency of arguments already developed by the student;
- point out possible logical flaws or contradictions;
- generate counterarguments that the student has to rebut;
- suggest additional references that will later be critically checked.
In this approach, AI becomes a kind of mirror and intellectual sparring partner, helping refine original ideas instead of supplying them ready-made. The student remains the author, but gains a partner to stress-test and expand their own reasoning. The key concern, as specialists emphasize, is to avoid using the model to generate the core ideas themselves – or even to suggest who to vote for, which some students reportedly asked, and which researchers see as very troubling.
Long-term risks for diversity of thought
The main concern of the scholars featured in the story is not just about today’s assignments, but about the cumulative effect of this dependence on a generation that is learning how to think. If, for years, students outsource reading, summarizing, analysis, and even decision-making, there is a real risk that they will lose the habit of sustained cognitive effort. And without that mental muscle, it becomes much harder to innovate, challenge dominant ideas, and criticize ready-made narratives, including political ones.
There is also a feedback loop at play. The more people use AI to write texts, create summaries, and structure arguments, the more those outputs end up back on the internet and, over time, in the training data for the next generation of models. This tends to further increase the homogenization of language and perspectives, in a cycle where the machine learns from its own output and keeps reinforcing its preferred patterns.
Socially, researchers warn about the risk of a less diverse intellectual culture, in which unconventional ways of thinking, marginal writing styles, and minority worldviews are pushed away from the center, not necessarily through explicit censorship, but through a mix of technological convenience and mental laziness.
Possible paths to preserve creativity
If banning AI from educational settings is not realistic, the challenge becomes finding ways to live with it without sacrificing originality. Some of the paths under discussion include:
- Designing questions that require personal experience: assignments that demand connections with concrete experiences, local contexts, or individual journeys are less vulnerable to generic answers.
- Valuing the process, not just the product: requiring drafts, reasoning logs, intermediate versions, and reflections on how the work was done helps differentiate those who truly thought things through from those who only polished a pre-made text.
- Teaching metacognition: encouraging students to observe how they think, which strategies they use, where they usually get stuck, and how AI fits into that flow, instead of running the tool on autopilot.
- Exploring other forms of expression: beyond written text, giving room for oral presentations, prototypes, maps, videos, and hands-on projects, where personal voice and improvisation still carry a lot of weight.
At the same time, there is an expectation that tech companies will invest in models capable of reflecting a broader diversity of styles, cultural contexts, and ways of reasoning. Until that happens at scale, the greater responsibility falls on schools, universities, and, of course, on the students themselves, who need to decide when it makes sense to use AI and when it is better to face the discomfort of thinking on their own.
In the end, the core point running through this whole debate is simple but heavy: delegating everything to the machine may make the text prettier, but it comes with a high cost in terms of intellectual autonomy. Technology can be a powerful partner, as long as it does not become the autopilot of our thinking. The hard work – and also the most rewarding – is still about building your own ideas, stumbling, revising, debating, and improving. That is where the richest part of education lives, and it is exactly there that creativity needs to stay alive, even in a world flooded by giant language models.
