UC Berkeley professor admits to using AI to edit op-ed on students' math skills
UC Berkeley math professor Zvezdelina Stankova admitted to using AI to help edit an op-ed published in the San Francisco Standard, which argued that test-blind admissions have caused severe math deficiencies among students The Daily Californian flagged 33% of the piece as AI-assisted using Pangram detection software, sparking public debate about AI use in academic writing Stankova defended her use of AI as limited to editing and research assistance, emphasizing the article resulted from "several
Analysis
TL;DR
- UC Berkeley math professor Zvezdelina Stankova admitted to using AI to help edit an op-ed published in the San Francisco Standard, which argued that test-blind admissions have caused severe math deficiencies among students
- The Daily Californian flagged 33% of the piece as AI-assisted using Pangram detection software, sparking public debate about AI use in academic writing
- Stankova defended her use of AI as limited to editing and research assistance, emphasizing the article resulted from "several hundred person-hours of intensive human work"
- The incident highlights growing tensions around AI detection reliability, with UC's AI council member noting most tools remain "quite unreliable and lack nuance"
- The controversy intersects with an ongoing debate about reinstating standardized testing at UC, which the academic senate has agreed to review for potential fall 2028 admissions changes
Why It Matters
This case illustrates the growing friction between academic professionals and AI detection tools, raising questions about transparency, authorship, and the reliability of tools like Pangram that claim near-perfect accuracy. It also reflects a broader institutional struggle to establish clear AI usage policies in academic and editorial contexts, as detection technology outpaces policy development.
Technical Details
- Pangram AI detection software flagged 33% of the 2,000-word op-ed as AI-generated or AI-assisted, though Pangram claims a 99.66% accuracy rate with one false positive per 24,000 documents
- Stankova disclosed using AI for editing assistance and for locating research documents, while maintaining that all substantive analysis was produced by human researchers
- The UC system operates an "AI council" to develop institutional AI initiatives, but no specific guidance currently exists for faculty AI use in personal research or editorials
- The op-ed presented data claiming students with severe calculus I readiness deficits tripled after test-blind admissions were introduced in 2020
Industry Insight
- Academic institutions need to establish clear, nuanced AI usage policies before detection tools become a primary enforcement mechanism, as current tools lack the reliability needed for high-stakes accusations
- The incident demonstrates how AI detection can shift focus from substantive arguments to procedural controversies, potentially undermining legitimate discourse on important policy issues
- Publishers and editorial organizations should proactively define AI assistance boundaries rather than reacting to individual cases, as the Stankova incident shows the reputational risks of ambiguous policies
Disclaimer: The above content is generated by AI and is for reference only.