AI Cheating: Study Shows ChatGPT Surpasses Harvard Students
A recent study has revealed that OpenAI’s language model, ChatGPT, is capable of outperforming Harvard students in coursework. The experiment was carried out by a second-year student at Harvard, who was concerned about the potential over-reliance on AI technology and its impact on critical thinking and writing skills.
The experiment involved submitting essays written by the student and ChatGPT to eight professors and teaching assistants for grading. Surprisingly, the results demonstrated that ChatGPT was able to obtain a GPA of 3.34, receiving mostly As and Bs with just one C. While this GPA may not be stellar by Harvard standards, it is significantly higher than a passing grade.
This experiment poses a significant challenge for educators who are struggling to keep up with the rapid adoption of AI tools that are difficult to detect and capable of producing college-level work. Professors are now reconsidering assignments and developing guidelines to establish boundaries between acceptable usage of AI tools like ChatGPT and cheating. Furthermore, it highlights the potential threat artificial intelligence poses to entry-level jobs in various sectors, including consulting, banking, journalism, and software engineering.
The increasing presence of AI tools in universities has raised fundamental questions about the future of education. Max Tegmark, a professor at MIT researching artificial intelligence, expresses concern about determining which skills to teach students that will remain economically valuable by the time they graduate.
Paul LeBlanc, president of Southern New Hampshire University, believes that the disruption caused by AI can be an opportunity to prioritize human-centered jobs such as teaching, counseling, social work, and healthcare. He argues for a shift in focus and investment towards fixing broken social systems in America.
As universities grapple with the rise of AI, Harvard has taken steps to address the issue. Harvard’s dean of undergraduate education, Amanda Claybaugh, recently shared guidance with faculty on incorporating generative AI into the classroom. Individual instructors will have the autonomy to draft policies for their classes regarding the usage of AI tools like ChatGPT. For example, some instructors may permit students to use language learning models for brainstorming and gathering inspiration, but restrict their use for actual essay writing.
Adriana Gutiérrez, a senior preceptor at Harvard, emphasizes the importance of students acknowledging their use of AI tools in assignments while prohibiting reliance on them for editing, translating, or writing essays. Gutiérrez highlights the advantage of small classes where instructors can identify discrepancies in writing style, allowing them to detect instances where AI tools have been utilized.
Some professors, like economics professor Jason Furman, have chosen to revise their assignment structures due to the emergence of ChatGPT. Furman intends to drop essays from Harvard’s introductory economics courses in light of the reduced net benefit resulting from the use of AI in essay writing. Instead, he plans to explore alternative assignment formats that challenge students to critique essays generated by ChatGPT.
Harvard will be organizing information sessions next month to assist professors in creating assignments that are impervious to AI-generated content. This proactive approach aims to address the concerns raised by Bodnick’s experiment and ensures that the boundaries between human work and AI assistance are clearly defined.
The prevalence of AI tools like ChatGPT raises important questions about the future of education and necessitates a reevaluation of pedagogical approaches. Acknowledging the inevitability of AI’s presence in various industries, universities must adapt to its integration while emphasizing the importance of critical thinking, creativity, and the writing process.
In conclusion, the rise of AI tools in education prompts both challenges and opportunities. While AI technology can enhance efficiency, it also necessitates careful consideration to prevent overreliance and ensure the preservation of critical skills. As universities navigate this new landscape, it is essential to strike a balance that embraces the benefits of AI while upholding the value of human-centered education.
