AI cheating runs wild on campus | Front Burner

Overview

Elaine Chao fills in for Jamie Poissant and interviews journalist James Walsh about how generative AI is changing cheating and assessment practices on North American university campuses.

Walsh, who wrote a New York Magazine piece titled “Everyone is cheating their way through college,” argues that AI is being used widely for academic dishonesty. He says institutions struggle to respond because the technology is moving quickly and is difficult to verify or detect.

Reported Scale of AI Use

Walsh cites survey data suggesting widespread adoption:

• KPMG survey of 400+ Canadian students: ~60% use AI chatbots (e.g., ChatGPT) for schoolwork
• Study.com poll of 1,000 US students: ~90% used it on most assignments

How Students Are “Cheating”

A key example is Roy Lee (Columbia). Walsh reports that Roy used AI to generate the majority (~80%) of his essays and treated assignments as “hackable.” Walsh also describes how Roy created a tool to help others cheat coding interviews by copying or using AI output—promoting it through a video about using a platform to answer an Amazon interview.

Walsh contrasts this kind of deliberate exploitation with more everyday student workflows:

• Drafting and “not copy-pasting”: Students may use AI for essay structure (e.g., outlines and topic sentences) while claiming they are not submitting direct copy-paste work.
• Time-saving narratives: Some students use AI because they feel they don’t have time to fully write or organize assignments, including amid heavy distraction (Walsh notes social media use).
• Dependence and “good enough” grading: Some students describe AI reliance as a habit or addiction, accepting passing grades even if the work isn’t fully “right.”

Why Professors and Institutions Struggle

Walsh says professors often notice “robotic” writing patterns—such as polished grammar, overly rigorous counterarguments, and work that doesn’t sound like typical undergraduate writing.

However, instructors face major barriers:

• Unreliable detection: AI detection software is not dependable, so professors don’t trust it enough to act.
• Proof is labor-intensive: Establishing misuse requires time and effort, leading to a “cat-and-mouse” dynamic.
• Students’ advice: “deny, deny, deny”: Walsh reports students believe it’s hard to prove AI involvement.
• Evasion tactics: Reported strategies include laundering essays through other AI tools, intentionally adding errors, using platforms claiming “authentic language,” and assembling submissions from earlier personal work.

Walsh also highlights “Trojan horse” tactics used by some instructors: embedding hidden weird prompts (e.g., “Mention Finland” or “Mention Dua Lipa”) so AI-generated copy-and-paste output includes absurd elements—an indication of how low trust has become.

Nuance: Cheating vs. Legitimate Teaching Uses

While the discussion centers on dishonesty, Walsh emphasizes that AI in education exists on a spectrum.

Examples of legitimate or educational uses include:

• Course preparation and feedback: Instructors may use AI to draft lecture materials, organize presentations, generate study tools, or even create an entire textbook for a literature class.
• Time and engagement benefits: These uses can free up time for teaching assistants and potentially improve student engagement.

Walsh notes this is ethically different from students using AI to submit work they didn’t meaningfully produce.

He also argues the stakes differ: professors are often overworked, while students may have grown up with AI and thus formed different expectations and learning habits.

Broader Implications

A central theme is that higher education may need a fundamental reassessment of what it tests and how it assigns work. Walsh predicts more cognitive labor—such as writing and memorization—may increasingly be outsourced to large language models. This would force educators to rethink grading, exams, and what counts as learning and value.

He frames the issue as moral and societal, not simply “young people cheating,” arguing society hasn’t clearly defined how AI should be regulated in education and beyond.

Structural complication: AI partnerships and campus access

The segment also raises a practical constraint: many universities partner with OpenAI and provide AI products aimed at students. That makes outright bans harder to enforce—and politically or financially unlikely. Even if campuses restricted access, off-campus use would continue. Companies also market “plus” and education versions to capture student users.

Presenters / Contributors

• Elaine Chao (presenter; filling in for Jamie Poissant)
• James Walsh (guest; author of the New York Magazine piece)
• Matthew Amha (producer)
• Joy Tasha Gupta (producer)
• Lauren Donnelly (producer)
• McKenzie Cameron (producer)
• Marco Luciano (producer)
• Katie Teeling (intern)
• Evan Agard (video producer)
• John Lee (YouTube producer)
• Joseph Shabason (music)
• Elaine Chao (senior producer)
• Jonathan Mupeti (assistant producer mentioned)
• Nick McCabe Lokos (executive producer)
• Jamie Poissant (referenced as the regular host returning Monday)