Study Finds AI Tools Are Inflating Student Grades, Pointing to Outsourced Work

A recent study from UC Berkeley indicates that the widespread adoption of AI tools like ChatGPT may be artificially inflating student grades, suggesting a shift towards outsourced work rather than genuine learning advancements. The research analyzed over 500,000 grades from a large, selective public research university in Texas.

The findings reveal a notable increase in student grades in courses that heavily rely on writing and coding assignments since ChatGPT became widely accessible in November 2022. Specifically, the proportion of ‘A’ grades saw a jump of approximately 13 percentage points, representing a roughly 30% increase compared to the pre-ChatGPT baseline of 2022. The average GPA also rose by 0.12 points, with a noticeable narrowing of the grade distribution, particularly an increase in straight ‘A’ grades at the expense of ‘A-minus’ and ‘B-plus’.

Study Details and Methodology

The study, led by Igor Chirikov, examined grade trends across 319 courses spanning 84 departments over eight fall semesters, from 2018 to 2025. To assess the impact of AI, courses were categorized based on their assignment mix from fall 2022 syllabi, before the widespread availability of AI tools. The core research question was whether these grade increases reflected actual learning gains or simply the completion of work by AI.

A key aspect of the study involved analyzing the weight of homework in the final course grade. Chirikov hypothesized that if AI were genuinely enhancing learning, grade improvements would be evident regardless of assessment methods, whether homework or proctored exams. However, the data indicated that the impact of AI was concentrated in courses where homework constituted a larger portion of the overall grade. In these courses, ‘A’ grades increased by an additional 16 percentage points compared to courses with lower homework weight but similar AI exposure. Conversely, in courses with less reliance on homework, the grade inflation effect was minimal and not statistically significant.

Chirikov’s findings suggest that the observed grade inflation is “difficult to reconcile with broad learning gains or sorting effects alone.” A placebo test focusing on oral presentation assignments, where AI is less effective, showed no significant change in grades, further supporting the conclusion that AI’s impact is tied to its capabilities in producing written and coded content.

AI’s Unique Impact on Education

While grade inflation is not a new phenomenon in US universities, with factors like institutional policies and evolving teaching evaluations contributing to rising marks historically, AI introduces a different dynamic. Unlike previous drivers that influenced grading after work submission, AI fundamentally changes how academic work is produced before instructors even see it.

The study warns that if grades increasingly reflect AI-generated output rather than genuine student skills in writing and coding, it could lead to flawed selection decisions by employers and graduate programs. Furthermore, Chirikov flags a potential feedback loop: if AI-driven tools replace skill-building tasks during college, graduates might be less proficient in areas where AI excels, potentially accelerating automation and widening skill gaps in the job market.

Potential Solutions and Future Outlook

To address these challenges, the study proposes rethinking assignment formats. While moving entirely to proctored exams is not a simple or sufficient solution, designing assignments that either limit AI use or integrate it intentionally, perhaps through documentation of the work process or follow-up assessments that verify true understanding, could be more effective.

OpenAI CEO Sam Altman has previously acknowledged the slow response of the education system to AI, expressing concern that critical thinking skills risk significant atrophy without systemic change. He believes adaptation is possible, as it has been with previous technological shifts, but emphasizes the continued importance of foundational skills like writing and coding for cognitive development.

Key facts

This development is crucial for ReviewArticle’s readership as it highlights a significant challenge in the current educational landscape directly influenced by AI technologies. The study provides concrete data on how AI is reshaping academic assessment, with implications for the perceived value of degrees and the skills of future graduates entering the workforce. Understanding this trend is vital for professionals in AI, education, and industries reliant on skilled human capital.

Source: The Decoder – AI is inflating student grades, and the effect points to outsourced work, not better learning – https://the-decoder.com/ai-is-inflating-student-grades-and-the-effect-points-to-outsourced-work-not-better-learning/