Since launching in November 2022, OpenAI’s ChatGPT, a generative artificial intelligence, has been compared to “steroids” numerous times. “ChatGPT is like steroids for your skills,” says one Reddit user. It’s “Google on steroids,” claims a journalism professor. AI will be like Photoshop but “on steroids,” OpenAI CEO Sam Altman told a U.S. Senate subcommittee last year.
Perhaps “steroids” is the right term. Like steroids in sports, generative AI could create a global cheating crisis that undermines the purpose and value of an education.
International sporting had its ChatGPT moment in 1998 when French customs agents found narcotics, testosterone, amphetamines, growth hormones, and syringes inside the car of Willy Voet, the caretaker of the Festina cycling team, on his way to the Tour de France. Further investigation revealed that the entire Festina team took banned substances in coordination with their management and doctors. All three medalists in the 1998 Tour de France were later found to have been doping.
Subsequent doping scandals in Major League Baseball and other major athletic leagues further tarnished the reputation of sports as a fair and meritocratic institution.
The rise of “AI doping” is strikingly similar. In a Study.com survey of 1,000 college-age students, 89 percent of respondents admitted to using ChatGPT to complete a homework assignment. Another 48 percent admitted to using it on at-home tests or quizzes, and 53 percent had the bot write an essay. Yet, 72 percent of the students reported believing ChatGPT should be banned from campus networks.
Are they hypocrites? No. When enough players in a competitive game can cheat with a high upside and low risk of consequences, other players will feel forced to cheat as well. As Lance Armstrong told the French newspaper Le Monde in 2013, several months after publicly admitting he had used performance-enhancing drugs, it was “impossible to win the Tour de France without doping.”
If enough students improve their grades using ChatGPT, their peers may conclude that it’s “impossible” to compete unless they cheat, too. In cycling, not doping meant losing competition winnings and sponsorships. In education, not using generative AI could mean losing out on college admissions, scholarships, and career opportunities.
At the height of the sports doping crisis, a common argument was to let doping happen. If no substance is banned, isn’t the playing field level? The counterargument is that substances are banned because they pose a health risk to athletes. If sporting organizations not only allowed but tacitly encouraged athletes to dope, the resulting biochemical arms race would have a sure loser: athletes and their well-being.
Likewise, the normalization of AI doping would create an arms race among students, resulting in several consequences for them and society:
1. Unchecked use of AI renders education pointless. We fund public education as a common good because it empowers citizens to live fulfilling lives and contribute to their communities. If the point of education is merely to get a diploma, then who cares if a student or AI does the work? The true point of an education, however, is to train a person’s mind and character.
Claiming that students no longer need to learn skills like writing—because AI does it—is like arguing that no one should strength train because carts and forklifts move heavy stuff for us. This conflates means and ends. People lift weights for the inherent benefits to their mind and body. Likewise, we learn to write for the inherent benefits to our cognition and communication skills.
2. AI threatens to undermine academic integrity, the foundation for professional credibility. We trust our surgeon, certified public accountant, or lawyer because we trust the institutions that test, certify, and employ them.
Well, researchers have found that ChatGPT can pass the U.S. medical licensing exam, CPA exam, and bar exam. You might argue that cheating on those exams is almost impossible. But what about an online nursing exam, an online certification in cybersecurity, or an online degree in social work? A person who cheats for the credential in those cases could become a danger to others.
3. The struggle to maintain academic credibility could produce a two-tiered education system that is even more inequitable than the current one. Elite colleges with full-time professors and graduate students have the resources to design assignments in which AI provides no edge. Community colleges and online education platforms don’t have that luxury. A class with hundreds of students and one part-time instructor cannot convert every digital test into an original research project or in-person test with pencils and paper—not without raising costs considerably.
Efforts to democratize education will be laughable if the only credible degrees come from private, in-person institutions that cost students an average of almost $56,000 per year.
K-12 schools will experience similar inequities, particularly between expensive private schools and crowded public schools. Schools with lower student-teacher ratios are better positioned to design assignments that limit students’ reliance on AI than schools with more limited staff and resources.
So how do we address AI’s threat to academic integrity and an affordable education?
Forget watermarking AI-generated text and AI detectors—they’re easily duped. And forget academic “honor codes.” In my birth country of Afghanistan, the strict honor codes of Islam that forbid corruption didn’t stop the country from becoming an epicenter of corruption after the United States’ 2001 occupation injected billions of dollars. Like Lance Armstrong, no one struggled to justify corrupt behavior when everyone else was doing it, too.
Exams, whether in person or online, must be proctored such that no one can cheat using AI. That said, to prepare students for the working world, schools should teach generative AI in classrooms using versions with limited capabilities.
The silver lining of the 1998 Tour de France was that the International Olympic Committee formed the World Anti-Doping Agency the following year. Though far from perfect, WADA created a unified list of banned substances and standards for detecting them. In other words, the organization defined what “doping” means in sports. Soon enough, education systems may need a WADA-like organization to define cheating in the AI age and set standards for preventing and detecting it.
Doping in sports undermined the fairness and meritocracy of a beloved institution, until that institution took the threat seriously. It’s time we take AI doping in schools seriously.