Teach for the Fog: Why Higher Ed Matters More in the AI Era
I strongly disagree with the advice to skip college and jump straight into work. Learning by doing is valuable, but it's most effective when people know which skills will hold value. The labor market is uncertain. Betting a career on today's job ads is shortsighted.
As one AI pioneer put it, progress feels like driving through fog: you can see a few feet ahead, not the next turn. Our job in education is to prepare students to operate in that fog. Train for adaptability, not a fixed checklist. We need drivers who can handle unfamiliar roads and surprises.
What to Teach: Durable Thinking, Not Disappearing Tasks
Go back to basics. A liberal education teaches how to think: reason well, read closely, write clearly, and judge evidence. These skills outlast software updates and changing tools. Students should use AI-critically. The goal is discerning users and capable judges, not passive consumers.
Keep teaching math, logic, and argumentation. Expose students to foundational texts and the structure of good thinking. Show how claims are built, tested, and revised. That's how people stay ahead of fast tech cycles.
AI Changes the Work-So Change the Assessment
LLMs are strong at summarizing, extracting key ideas, coding scaffolds, quantitative steps, and even drafting prose. These activities stay in the curriculum, but the goal shifts. Students need concept mastery and logic, not rote execution. Success will look like setting clear goals, structuring problems, and using tools wisely.
Assessment must adapt. Homework essays, take-home problem sets, and unsupervised exams are weaker signals now. Replace them with in-person exams, oral defenses, whiteboard work, and timed studio tasks. Understanding still needs practice, but we have to see the student do the thinking.
Practical Moves You Can Make This Term
- Set AI-use rules by task. Require disclosure of prompts and outputs. Ask for a short "how I used AI" memo with every submission.
- Grade the process. Collect drafts, version history, notes, and reflection. Do quick viva checks on randomly selected students to confirm authorship.
- Bring work into the room. Use whiteboard sprints, coding labs, case crunches, and mini oral quizzes. Short, frequent, live checks beat one big take-home.
- Teach reasoning directly. Logic drills, argument maps, estimation, probability, and causal inference. Less grind, more thinking.
- Use AI as a foil. Have students compare model outputs, critique errors, and improve them. Always ask: what evidence supports this, and what's missing?
- Update rubrics. Score problem framing, assumptions, selection of methods, and justification of tool use-not just final answers.
- Reduce bias risk. Use clear criteria, two graders for high-stakes orals, and brief recordings when appropriate. Provide students with sample performances.
- Protect equity. If you can't shrink every class, add recitations, peer review, and rotating small-group orals. Use AI for rapid formative feedback, then spend human time where it matters.
- Upskill faculty. Run short workshops on AI literacy and assessment redesign. If you need structured options, see AI course paths by job.
Curriculum Guidance for the AI Era
Scope less "how to do each step by hand," more "why this step, and when." Keep programming, stats, and writing, but push students to explain choices, check outputs, and cite sources. Treat AI like calculators and spellcheck: useful helpers, not replacements for thought.
Blend domains: data literacy + ethics + communication. Make cross-course projects that require problem scoping, constraints, and resource planning. That's the skill set work rewards.
What to Measure
Measure transfer: can students apply ideas in new settings? Measure problem definition, prioritization, and use of tools under time and resource limits. Make students state their goal, plan, and why an AI is the right tool for this job.
The Cost, the Risk, and the Fix
Smaller, face-to-face classes are expensive. Elite schools can pivot; big publics will feel the strain. Without support, gaps will widen. Institutions need to back faculty with standards, training, time, and staffing.
AI will replace jobs and create new ones. Education should grow, not shrink. Quality will depend on expectations and enforcement. With AI productivity gains, funding smaller sections, more instructors, and more personal contact is both possible and worth it.
Further Reading
Your membership also unlocks:
