Search

Cookies

We use cookies to improve your experience. By continuing, you accept our use of cookies.

Technology

ChatGPT Eased Homework, But Grades Dropped 17%: Wharton Professor's AI Guide

· · 3 min read

Wharton Professor Ethan Mollick's new book, 'Co-Existence: The Next Phase of AI,' reveals that while AI can simplify tasks, students using ChatGPT for homework saw a 17% drop in grades. He argues that making learning too easy hinders genuine understanding and retention.

Artificial intelligence is rapidly advancing, but are we using it wisely? Ethan Mollick, an AI researcher and associate professor at the Wharton School, addresses this crucial question in his new book, 'Co-Existence: The Next Phase of AI.' Mollick provides a practical guide on how individuals can effectively integrate AI into their work and learning processes without compromising their critical thinking abilities.

The Hidden Cost of AI Assistance in Learning

Mollick emphasizes that AI is not a substitute for human thought. He points to a significant 'capability overhang,' where much of AI's potential remains untapped, yet its misuse can be detrimental. A study involving Turkish students from the University of Pennsylvania documented a stark example: students who used ChatGPT without restrictions for homework experienced a 17% decline in their grades compared to those who did not use AI. The professor explains, "They would ask the AI to help, and they would get a wonderfully detailed step-by-step approach to the solution. It felt like learning, but it wasn't."

This phenomenon, Mollick argues, stems from AI making the learning process too easy. Genuine learning requires effort and engagement to ensure retention, a principle often overlooked when AI provides instant solutions.

AI's Impact on Cognitive Engagement

Mollick also cites a 2025 MIT Media Lab study that explored the cognitive effects of AI use. This research found that students who employed ChatGPT to write essays were less engaged with the material and subsequently remembered less about their work. Four months later, these students performed worse when tasked with writing without AI assistance. The study further suggested, through brain scans, that students using AI showed reduced brain activity when recalling their essays, implying a diminished cognitive effort. While the methodology of the brain scan aspect was questioned by other scholars, the core idea that "AI damages your brain" resonated widely.

"AI does not physically damage your brain," Mollick clarifies. "But we latched on to the dramatic interpretation because it confirmed what we already suspected: that outsourcing our thinking to machines might atrophy our ability to think at all."

He likens this anxiety to historical concerns about new technologies, such as Plato's worry that writing would impair memory. While technologies like calculators have indeed led to the fading of certain mental skills, the issue with AI is more nuanced. The default setting for most AI, Mollick notes, is to act as a helpful assistant, providing direct answers rather than encouraging users to solve problems independently. This creates a problem where the user feels they are learning, but they are merely consuming pre-digested information.

Leveraging AI for 'Desirable Difficulties'

The solution, according to Mollick, lies in understanding the concept of 'desirable difficulties'—challenges that strengthen memory and understanding over time. When AI summarizes material or solves problems, it bypasses these essential difficulties, short-circuiting the learning process. Students in the Turkish study, despite reporting greater confidence and finding their work easier with AI, ultimately performed worse on tests.

However, Mollick is not against AI in education. He asserts that AI has a significant role to play, provided it's used mindfully. The negative impact observed in the Turkish study disappeared when AI was instructed to function as a tutor, guiding students rather than simply providing answers. A separate study in Taipei further supported this, showing that appropriately guided AI could actually improve students' test scores. The key takeaway is clear: using AI to avoid the hard work of understanding is self-defeating, but when leveraged as an interactive learning tool, its potential is immense.

Related