Berkeley Study Links AI to Grade Inflation
A University of California, Berkeley working paper titled "Artificial Intelligence and Grade Inflation" analyzes more than 500,000 grades from 2018 to 2025 at a large Texas research university and finds that courses exposed to generative AI saw the share of A grades rise by 13 percentage points, about 30% relative to the 2022 baseline, after the release of ChatGPT (per the CSHE working paper released May 13, 2026). The increases were concentrated in writing and coding courses and were larger where homework carried greater weight, which the paper interprets as consistent with AI substituting for students' take-home work. Igor Chirikov, the paper's author, said, "As much as AI is helping people become more productive... I think it may harm their learning." News coverage in outlets including the Wall Street Journal and The Chronicle summarized the findings and noted debate among instructors about assessment design.
What happened
The University of California, Berkeley Center for Studies in Higher Education released a working paper, "Artificial Intelligence and Grade Inflation," on May 13, 2026. Per the paper, the author analyzed more than 500,000 student course enrollments across 319 courses in 84 departments at a large, selective Texas research university between 2018 and 2025 and used a difference-in-differences design to estimate effects after the release of ChatGPT. The paper reports that courses classified as more exposed to generative AI, especially writing and coding classes, experienced an increase in the share of A grades of 13 percentage points, roughly 30% relative to the 2022 baseline (CSHE working paper, Igor Chirikov).
Technical details
Per the CSHE working paper, the author exploits variation in course task types and the weight of homework versus in-person assessments, finding larger grade increases where homework carried more weight. The paper reports robustness checks intended to separate AI-driven substitution from alternative explanations such as AI-enabled tutoring or course selection, and cites the homework-weight pattern as evidence consistent with AI performing graded tasks (CSHE working paper).
Context and significance
Editorial analysis: Reporting across outlets including the Wall Street Journal, The Chronicle of Higher Education, University World News, and news aggregators contextualizes the paper within ongoing debates about assessment validity, academic integrity, and credential signaling. The study's numeric scale and design have made it a focal point for discussions about whether observed grade rises reflect learning gains or easier achievement via tools like ChatGPT (news coverage; CSHE working paper). Chirikov warns that higher grades may reflect access to powerful tools or skill at using them rather than underlying ability, a point he raises in quoted remarks to the press.
What to watch
For practitioners: indicators to follow include replication studies using different institutions, changes in course assessment mixes (homework weight versus invigilated exams), campus adoption of AI-detection or honor-code measures reported by instructors, and employer responses to credential signals. Observers should also track whether subsequent research corroborates the homework-weight mechanism across more heterogeneous institutional settings.
Scoring Rationale
The paper uses a large dataset and causal-style design, making the finding notable for educators, employers, and assessment designers. The topic is immediately relevant to campus practice but is not a frontier technical advance, so the story rates as notable rather than industry-shaking.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
