Wiki › Thinking next to AI

AI Education With and Without Machines

UChicago curriculum sequencing: independent writing and reasoning before deliberate AI integration, with ethical and critical computational literacy.

Core idea

Paul Alivisatos, president of the University of Chicago, proposes an educational sequence rather than a choice between banning AI and using it everywhere: first develop independent writing, reasoning, and judgment; then deliberately integrate computational tools. Students should learn to think without machines, with machines, and critically about machines, becoming ethical, skeptical, and ambitious users of AI.

The distinction is between producing an answer and forming the person who can evaluate it. This connects AI and Mental Effort to institutional safeguards against Cognitive Surrender.

UChicago changes described on September 20, 2026

• Reform process: an interdisciplinary faculty working group began considering changes in fall 2024.
• Writing: a forthcoming required undergraduate writing course will exclude machine assistance, preserving drafting from a blank page and critique from instructors and peers.
• Humanities and social sciences: the core classrooms will exclude technology. This is not a university-wide prohibition on AI.
• Computational thinking: Alivisatos expects a new core requirement within the following year; this is broad support and a likely outcome, not a finalized requirement. The scope includes domain-specific computational tools and critical reasoning about how computation, data, and algorithms shape knowledge.
• Law school: the school announced that first-year students will learn writing and reasoning without classroom technology or AI, followed by deliberate AI integration in years two and three.

Why the sequence matters

The intended mechanism is foundational apprenticeship before tool-assisted acceleration. Protected unaided work develops the judgment needed to scrutinize later machine-assisted results. AI literacy also includes authorship, ethical use, and legal and democratic governance—not merely prompting skill.

This parallels Code Reading as Cognitive Labor: work can be valuable because it builds expertise, even when a machine could produce the immediate artifact faster. This connection is wiki synthesis, not a result demonstrated by the article.

Evidence limits and open questions

The source is an opinion column and a first-person account of planned reforms, not a comparative evaluation of educational outcomes. It offers no measured evidence that this sequence produces better learning than alternatives. Faster advances in knowledge and well-being are the author's aspiration.

Open questions: how to assess unaided capability, when students are ready for AI integration, how to distinguish formative effort from unnecessary friction, and how to verify compliance outside the classroom.

Source

“Teach Students to Think With, Without and About AI,” Paul Alivisatos, The Wall Street Journal, September 20, 2026. Captured from the full article text supplied by Andrey; direct WSJ access was blocked by device verification.