AI Resilience Report forPhysics Teachers, Postsecondary

Right now, AI is mostly augmenting — not replacing — postsecondary physics teachers. It's showing up as a helpful sidekick for the tedious parts of the job. A 2025 NPR-reported Anthropic study of ~74,000 higher-ed conversations found the majority — or 57% of the conversations analyzed — related to curriculum development, like designing lesson plans and assignments, and that professors tend to automate more tedious and routine work, including financial and administrative tasks, while teaching and lesson design was much more of a collaborative process.

Grading is a mixed story: about 7% of conversations were about grading, and when educators use AI for grading, they often automate a lot of it away — but faculty said grading was the task the chatbot was least effective at. In physics specifically, the American Association of Physics Teachers now runs an ongoing "AI Physics Tools" column in The Physics Teacher ^[1] with 2026 articles on using DeepSeek for dynamic visualizations, generative AI for free-fall experiments, and ChatGPT-Phyphox harmonic analyses. The American Physical Society also published a 2026 focused collection on AI tools in physics teaching ^[2] covering assessment, grading, and adoption.

So the "high-automation" tasks in your list — attendance/grade records, exams, and syllabi drafting — are exactly where AI is landing, while supervising research and serving on committees remain deeply human.

Adoption is happening fast, but with real friction. The Digital Education Council's 2026 Global Survey ^[3] of 45,398 people across 35 countries found that 77 percent of faculty use it in their teaching, up 16 percentage points on 2025, driven by cheap, commercially available tools like ChatGPT, Claude, and Gemini. But adoption is uneven — in the US and Canada, faculty intent to use AI in teaching has fallen nine percentage points in a year, from 76 percent to 67 percent, the lowest of any region, and 55 percent of faculty say AI poses a serious risk to human intellectual development.

Ethical worries are a major brake: an Inside Higher Ed essay on STEM classrooms ^[4] describes physics-style problem sets where ChatGPT produced "confidently, elaborately wrong" answers, showing why physics professors are cautious. Institutional readiness also lags — the 2026 EDUCAUSE Impact of AI on Learning Assessment report ^[5], based on 438 faculty and staff, highlights "growing momentum" alongside "real uncertainty" about integrity and policy. The good news for you: the parts of teaching that require judgment, mentorship, and human connection — running labs, guiding research, and coaching students through hard concepts — are exactly the parts AI is worst at, and are exactly what make a great physics professor irreplaceable.