Yale University senior Amanda always knew that many of her classmates were using AI chatbots to write essays or do homework. But in small seminars, she started to notice something really strange: her classmates would sit at their laptops with perfectly polished abstracts, but the live discussions after their presentations would simply fail.
Yale University senior Amanda always knew that many of her classmates were using AI chatbots to write essays or do homework. But in small seminars, she started to notice something really strange: her classmates would sit at their laptops with perfectly polished abstracts, but the live discussions after their presentations would simply fail.
The dev.ua editorial team retells a CNN article about how AI has already impacted education in the United States.
«Somehow the conversation got to a dead end, I looked to my left and saw a guy typing furiously on his laptop, asking the chatbot the question the professor had just voiced,» says Amanda.
She and two other Yale students (names changed for privacy reasons) shared their observations: The widespread use of AI in classrooms is destroying discussion. «Everyone sounds the same now,» Amanda says. «Before, everyone had something unique, approached the topic from different angles. Now, that’s gone.»
As artificial intelligence becomes more deeply integrated into our lives, educators and scientists are sounding the alarm: the technology may gradually destroy students' ability to think original thoughts.
Student Jessica admits to using AI every day. In economics seminars, where the professor can call on anyone, she sees how at the beginning of a pair «everyone uploads all the PDFs of the materials» into a chatbot.
She also uses AI when she’s having trouble coming up with ideas. However, Jessica admits to a sad fact: chatbots have made her lazier, and her work ethic has «completely degraded» compared to her high school days.
Professor Thomas Chatterton Williams of Bard College notes an interesting paradox. On the one hand, AI has «raised the baseline» for group discussions on complex concepts. On the other hand, it has killed off the emergence of strange, unconventional, and original ideas.
«My biggest concern is that many smart young people will never gain their own voice,» Williams concludes.
Large language models (LLMs) are trained to predict the next most statistically likely word. As Zhivar Surati, a doctoral student at the University of Southern California (USC), points out, this poses a serious problem.
In March, a study was published in the journal Trends in Cognitive Sciences, which proves that LLMs systematically homogenize (make homogeneous) human self-expression along three dimensions: language, perspective (point of view), and reasoning strategies.
The data on which the models are trained is oversaturated with dominant ideas. Therefore, AI often reproduces so-called «WEIRD» views (Western, educated, industrialized, rich and democratic).
When a group of people constantly interacts with AI, their creativity becomes «flat» compared to those who work without the help of algorithms. Instead of a variety of thoughts to open-ended questions, teachers receive perfectly polished but banal answers reduced to a few categories.
Daniel Buck, a fellow at the American Enterprise Institute and former teacher, calls AI «a supercharged version of SparkNotes» (a popular site for short book summaries). But where before a teacher could easily catch a student who had only read a summary, now AI can generate a unique answer to any question.
The problem is that students avoid the so-called «cognitive struggle.»
«A lot of learning happens in the nitty-gritty, in the grappling with the material,» Buck explains. «If a student delegates thinking to an AI, they might be able to replicate a thesis in class, but they won’t gain the basic skills to apply that knowledge elsewhere.»
USC study co-author Morteza Deggani goes even further, warning that intellectual laziness can have a societal impact, from a loss of innovation to an inability to critically evaluate political candidates.
Higher education institutions are trying to adapt. At Yale University, for example, AI policies are left to faculty discretion because generated text detection tools have been found to be unreliable.
To combat «cheating» and make students think, teachers resort to radical (or well-forgotten old) steps:
AI is not going anywhere, and experts advise changing the approach to its use. Student Sofia believes that it is better to honestly admit to the professor «I don’t understand the topic» than to read machine-generated «smart words» that are not backed by real experience.
Morteza Deggani suggests using AI as an opponent, not an executor: first write your own theses, and then ask the chatbot to find flaws in them.
«AI models should be collaborators. They should not be agents that do everything for us,» the scientist concludes.
