AI Over-reliance Hurts Student Skills

UC Berkeley CS professors are seeing a historic rise in failing grades, linking it to student over-reliance on AI for problem-solving.

Professors at UC Berkeley’s computer science department are reporting a significant increase in failing grades in foundational courses. Instructors have observed a noticeable decline in students' fundamental math, logic, and problem-solving abilities, which they attribute to an over-reliance on generative AI tools. According to faculty, students are using AI to generate complete solutions for assignments, allowing them to complete work without grasping the underlying concepts. This practice circumvents the critical learning process that builds a strong theoretical foundation, leading to poor performance on exams designed to test genuine understanding.

This development at a top-tier engineering school is a significant signal for the tech industry. Founders, CTOs, and engineering managers may soon face a hiring pool of new graduates who lack the core competencies essential for software development. While proficient at using AI tools, these candidates might struggle with debugging complex systems, reasoning through novel problems, or adapting algorithms to new contexts. The concern is that a generation of engineers could become overly dependent on AI, limiting their ability to innovate and solve problems that require deep, abstract thinking.

In response, educators are beginning to adapt their curriculum and evaluation methods. Some professors are redesigning exams to be in-person and focus more on conceptual understanding that cannot be easily answered by an AI. The challenge for both academia and industry is to find a balance where AI is used as a productivity tool that complements, rather than replaces, the development of fundamental engineering skills.

This trend at a top CS program suggests a potential decline in the fundamental problem-solving and critical thinking skills of the incoming tech workforce. Companies may need to re-evaluate hiring criteria and invest more in foundational training for new graduates.

Companies may face increased costs for training and onboarding new engineers who lack fundamental skills. Productivity could be affected if junior developers are unable to tackle novel problems without AI assistance, potentially slowing down innovation and increasing the burden on senior staff.