AI and the Future of Education: Teach Thinking, Not Tasks
Some claim college is on its way out. That misses the point. Work is uncertain, and training students for a fixed list of skills sets them up to fall behind the moment those skills shift.
The job is to build adaptable thinkers. People who can read closely, reason clearly, write with precision and judge evidence. Those abilities hold their value even as tools change.
Teach for the fog
One AI pioneer compared progress in this field to moving through fog - you see a few steps ahead, not the road. Prepare students to operate in that fog. That means fundamentals first and AI fluency second.
- Core math, logic and statistics: emphasis on reasoning, causality and proof over rote steps.
- Argument and writing: structure, evidence, clarity and revision; less emphasis on endless grammar drills.
- AI literacy: prompt planning, comparing outputs, citing sources, checking claims and detecting model failure modes.
- Concept over procedure: let tools handle grind work; students must explain the "why" and choose the right approach.
- Problem framing: define goals, constraints, trade-offs and metrics before touching a tool.
We've seen this before. Calculators and computers didn't end arithmetic or writing; they shifted time from manual steps to ideas. AI calls for the same adjustment. For context on technology in math learning, see this position from NCTM Technology in Teaching and Learning Mathematics.
Assessment that actually measures learning
AI makes shortcuts easy. That doesn't mean students learn less; it means we need assessments that surface real understanding.
- More in-person checks: short quizzes, whiteboard problem solving and oral defenses.
- Process over product: require plans, drafts, reasoning traces and citations of any AI use.
- AI-allowed, not AI-dependent: permit tools, but grade the clarity of goals, method choice and evaluation of outputs.
- Authentic tasks: messy data, conflicting sources and constraints that force judgment.
- Integrity by design: smaller stakes, higher frequency and randomized prompts reduce copy-paste incentives.
What this means for institutions
Stronger assessment leans on in-person time, smaller groups and more interaction. That's resource intensive. It also puts more responsibility on instructors to enforce standards and make calls.
- Support faculty: time, training, clear policies and backing on academic integrity.
- Guard against bias: shared rubrics, double-scoring when feasible, recorded oral exams and transparent appeals.
- Equity matters: small-class models are costly. Public institutions need funding so quality doesn't become a luxury good.
- Fund it with productivity gains: use AI to cut admin time and reinvest those hours and savings into teaching.
A simple playbook you can ship this term
- Publish an AI policy: what's allowed, what requires citation, what's off-limits and why.
- Redesign one assignment: require goal setting, tool choice rationale, output critique and a short oral check.
- Swap one take-home for a 15-minute in-class quiz or a 5-minute oral stand-up per student.
- Add an "AI lab" week: compare outputs from two models, verify claims with sources and reflect on failure cases.
- Train your team: a 90-minute workshop on prompts, verification and grading processes beats months of confusion. For structured options, see AI courses by job.
The bottom line
AI will replace tasks, then create new ones. Education should grow, not shrink. Smaller classes, more instructors and higher expectations cost money - and they're worth it.
As tools become widely available, quality will hinge less on access and more on standards and enforcement. Set the bar, teach the thinking and let AI handle the busywork. For broader context on AI literacy efforts, this overview from Stanford HAI is useful: AI Literacy.
Your membership also unlocks:
