AI is making students ‘similar to each other’

During her senior year at Yale University, Amanda learned that many of her classmates used chatbots to write essays and do homework. However, she only noticed something strange during the group discussion: the students put forward carefully crafted points and arguments, but the results were “bland and lifeless” on many different topics.

“During the class, I saw a familiar scene. When the lecturer asked students to think about a question, the friends next to them immediately typed furiously on their open laptops,” Amanda told CNNasked to use a pseudonym to avoid trouble.

“Nowadays, everyone discusses the same thing,” she said. “During first year, the debates were chaotic and incoherent, but everyone contributed something new, relied on each other’s opinions, approached from many angles and made different comments.”

In fact, according to an article published in March in the journal Trends in Cognitive Scienceslarge language models (LLMs) are “systematizing” human expression and thinking in three aspects: language, perspective, and reasoning.

In educational settings, students and faculty say they are seeing the effects of that trend in the classroom. And that causes many students to speak with similar views.

Impact

Jessica, a senior at Yale University, said she uses AI every day in class. “At the beginning of class, you can see each person putting a PDF file into the chatbot,” she said.

According to Jessica, AI’s ability to express ideas clearly is essential for her – who often has difficulty translating thoughts into words. “I want to comment, I have an idea, but I don’t know how to say it coherently,” she said. “So I asked a chatbot to make my sentences more coherent.”

Professor Thomas Chatterton Williams at the Hannah Arendt Center at Bard College, witnesses the impact of students’ decisions to use AI in many different directions. “Relying on AI has helped improve the quality of class discussions, especially in courses with difficult concepts,” he said. “But this technology also tends to discourage more strange, original and new thinking.”

According to him, the concern is that many talented young people will not have their own voice. Even a significant number of them will not truly understand the value of writing and having an opinion.

Jessica admitted that she found herself becoming lazier since starting to use chatbots in her studies. “It seems like my spirit of self-study and self-work has completely disappeared,” she shared.

Regarding the issue of AI making students’ accents sound similar, Zhivar Sourati, a doctoral student at the University of Southern California, said that large language models are trained to predict the next word most likely to appear based on the content that has appeared before. The answer therefore “reflects a narrow and distorted slice of human experience, a narrowing of the conceptual space that models use to write, speak and reason”.

Leading an ongoing study with a team of colleagues, Sourati explains that AI-induced homogenization occurs across three dimensions: language, perspective, and argumentation strategy. AI models tend to reproduce what is called the WEIRD (Westernized – educated – industrialized – wealthy – democratic) perspective. As a result, the AI ​​will favor the “more socially correct” WEIRD, overshadowing other viewpoints.

“When a person or group interacts repeatedly with an AI system, their creativity is reduced compared to without AI support,” Sourati wrote on the blog.

The “levelling” phenomenon raises concerns in educational institutions at all levels when applying AI. Morteza Dehghani, professor of psychology and computer science at the University of Southern California, a member of the research team, said this trend causes people to “lose diversity” in thinking, become intellectually lazy, and have a great impact on human society in the future.

Daniel Buck, a researcher at the American Enterprise Institute and former English teacher, is concerned that students are cognitively “avoiding” participating in class discussions and completing homework. “A lot of knowledge needs to be learned from small, boring details or from difficulties,” Buck said. “Students only remember what they actually consciously spend time on. If they entrust it to AI, they can only reproduce an argument, not build their own skills.”

This researcher is worried that if students rely too much on AI, they will graduate without building relationships with professors and persistent mental work habits. That is, they will have difficulty solving real-world problems when they graduate.

The problem is also related to cheating in exams. Last year, the use of AI to cheat occurred at many leading universities in Korea, such as Yonsei University, Seoul National University, Korea University… causing controversy and posing challenges in teaching and learning in the technology era, according to Korea Herald.

Respond

According to philosophy professor Sun-Joo Shin of Yale University, controlling and directing the use of AI for students is “a big task for anyone involved in teaching”. She emphasized that teachers/lecturers must continuously explore methods to ensure students continue to think critically and creatively in the age of artificial intelligence.

“I want my students to understand the content of the lesson – something that has not changed before and after the emergence of artificial intelligence,” she said. “I want them to use this exciting tool to their advantage, not become victims of it.”

Yale University now offers guidance on using AI for students and faculty on its website. The school encourages all instructors to tailor lesson plans to the specific course and learning goals of students instead of requiring cheat detection tools. Besides, they can control class lessons through surprise tests.

Danny Liu, professor of educational technology at the University of Sydney, thinks AI should not be banned in the classroom. “Instead of punishing, students should be taught correct usage. We want to verify whether students are learning, not whether they are cheating,” Liu told ABC late last year.

Educators emphasize that they can find ways to overcome students’ use of AI in assessments. However, it is equally important that students need to proactively limit their dependence on AI in the learning process.

Basil Ghezzi, a freshman at Bard College, said he actively avoids AI in studying, partly because this technology consumes resources and harms the environment. However, much of it lies in how AI has created “stereotyped friends” around.

“Talk to teachers, talk to professors, talk to people around you,” Ghezzi said about how to not be tempted by AI. “Create meaningful conversations with the people in your life.”

Professor Dehghani hopes more companies will invest in AI models that can reflect the diversity of thinking in society. However, for now, he believes people should limit their use of AI to generating ideas or making inferences in learning. “AI models should be collaborators, not agents who do things on our behalf,” Dehghani added.