Why generative AI 'hallucinates' and makes up stuff

University of Rochester’s Christopher Kanan says current iterations of AI lack human-like self-awareness and reasoning abilities.

Apr 10, 2025

2 min

Christopher Kanan

Generative artificial intelligence tools, like OpenAI’s GPT-4, are sometimes full of bunk.


Yes, they excel at tasks involving human language, like translating, writing essays, and acting as a personalized writing tutor. They even ace standardized tests. And they’re rapidly improving.


But they also “hallucinate,” which is the term scientists use to describe when AI tools produce information that sounds plausible but is incorrect. Worse, they do so with such confidence that their errors are sometimes difficult to spot.


Christopher Kanan, an associate professor of computer science with an appointment at the Goergen Institute for Data Science and Artificial Intelligence at the University of Rochester, explains that the reasoning and planning capabilities of AI tools are still limited compared with those of humans, who excel at continual learning.


“They don’t continually learn from experience,” Kanan says of AI tools. “Their knowledge is effectively frozen after training, meaning they lack awareness of recent developments or ongoing changes in the world.”


Current generative AI systems also lack what’s known as metacognition.


“That means they typically don’t know what they don’t know, and they rarely ask clarifying questions when faced with uncertainty or ambiguous prompts,” Kanan says. “This absence of self-awareness limits their effectiveness in real-world interactions.”


Kanan is an expert in artificial intelligence, continual learning, and brain-inspired algorithms who welcomes inquiries from journalists and knowledge seekers. He recently shared his thoughts on AI with WAMC Northeast Public Radio and with the University of Rochester News Center.


Reach out to Kanan by clicking on his profile.


Connect with:
Christopher Kanan

Christopher Kanan

Associate Professor of Computer Science

Christopher Kanan's research focuses on deep learning and Artificial Intelligence (AI)

AI and Machine LearningApplied Machine Learning (e.g. Medical Computer Vision)Language-guided Scene UnderstandingArtificial IntelligenceDeep Learning

You might also like...

Check out some other posts from University of Rochester

2 min

Don't let brain bias tank your fantasy football season

The National Football League season kicks off this week and that means millions of fantasy football coaches are already overthinking their lineups. But before they blame a bad draft slot or a fluke injury for bombing from one week to the next, they might want to look in the mirror and give their head a shake. Renee Miller, a professor of brain and cognitive sciences at the University of Rochester, studies cognitive biases and literally wrote the book on bias in fantasy sports. She plays fantasy football, too. She warns that our brains are wired to interpret fantasy football results in ways that are suboptimal and illogical. “Biased thinking occurs in everyday life and work, and in fantasy sports,” Miller says. “Through the course of a season, you can see a full range of the ways cognitive bias affects a person’s weekly fantasy matchups.” Here’s the good news: Miller says we can untangle those wires if we know what to look for. Among the biggest culprits are what Miller calls “the endowment effect” (overvaluing and clinging to players you drafted high), “recency bias” (falling in love with last week’s star), and “confirmation bias” (cherry-picking stats that support what you already believe). But especially beware of Week One. Thanks to the “primacy effect,” those games early in the season loom larger in memory than later ones. One hot debut or a disappointing flop can warp a coach’s thinking for weeks. The result? Lineups driven more by emotion than logic — and possibly a lot of pick sixes. Biases aren’t all bad, though. Sometimes instincts pay off. First impressions and recent performances sometimes hold fast. But the best fantasy players, Miller says, know when to slow down and think systematically. They stay skeptical, challenge their gut reactions, and accept that they’ll be wrong sometimes. So before you rage-drop that underperforming wide receiver or crown your Week One sleeper a superstar, remember, the smartest move might be to take a look in the mirror and give your head a shake. Miller is available for interviews for journalists covering fantasy sports. Connect with her by clicking on her profile.

1 min

Back-to-school stress? Here’s how it can be a good thing.

As America heads back to school, the renewed whirlwind of expectations for students and parents — from demanding coursework to social dynamics and balancing pick-up-and-drop-off schedules — can trigger anxiety for students and parents alike. Jeremy Jamieson, associate professor of psychology who leads the University of Rochester’s Social Stress Lab, studies how social stressors affect decisions, emotion, and achievement and how embracing, rather than battling, those reactions can boost resilience. “We’re not passive receivers of stress,” Jamieson told National Public Radio last year. “We’re active agents in actually making our own stress response.” Jamieson’s research reveals that stress can be helpful when it is reframed as a mobilizer of energy and focus. In a study of students preparing for the GRE, for instance, those who were primed to view physical stress symptoms (like a racing heart) as beneficial outperformed their peers who didn’t reframe those symptoms. As students confront the fall’s demands, a simple shift in mindset can make all the difference. Jamieson’s research has so many practical applications that he is regularly sought out by media outlets on a wide variety of topics. In the last year, he has talked to Golf Digest about battling the “yips,” to The Atlantic about the rise of “anxiety-inducing” television, and to New York Magazine about the stress some people feel when talking on the phone. He is available to discuss his research and to help explain and navigate seasonal pressures. Connect with him by clicking on his profile.

2 min

In an age of fast-moving misinformation, our expert teaches students how to spot what’s credible

As the new academic year begins, and at a time when misinformation often travels faster than facts, University of Rochester’s Kevin Meuwissen offers educators and young learners clarity and practical strategies for identifying credible sources. As an associate professor and chair of teaching and curriculum at the Warner School of Education and Human Development, Meuwissen focuses on how children and teens learn about politics and history — and how they can be taught to critically evaluate what they consume. “Young people pay close attention to who’s been consistently accurate,” he says. “They’re more likely to trust someone over time if their information holds up.” To empower students in our complex information environment, Meuwissen champions the so-called SIFT method — an easy-to-remember acronym and evidence-based toolkit that breaks down like this: • Stop! Pause before reacting or sharing • Investigate the source • Find better coverage • Trace claims back to their origin He also warns about how emotional framing, AI-generated visuals, deep fakes, and repeated exposure can distort judgment through the illusory truth effect — making misinformation feel believable even when it isn’t. His "Ever Wonder: How Can You Tell If A Source Is Credible?" video  is a handy teaching tool.  Meuwissen and his colleagues encourage teachers grappling with resistance over topics like climate science to consider not just evidence depth, but also students’ identities — political, cultural, and otherwise — when designing lessons. His approach emphasizes building trust, modeling thoughtful verification, and nurturing classroom norms rooted in accuracy — traits essential for forming discerning digital citizens. Kevin Meuwissen is available for interviews about identifying misinformation. He can be contacted through Warner School of Education Director of Communications Theresa Danylak at tdanylak@warner.rochester.edu.

View all posts