The Collision — book cover

The Collision

What AI Does to Us

Eric So

Published by W. W. Norton


AI is not just changing what we do, it is changing who we are as thinkers. This book examines the collision between artificial intelligence and human cognition: how our growing reliance on AI tools is quietly reshaping memory, judgment, learning, and even how we see the world around us. The pages are filled with practical frameworks for navigating this moment — understanding the pull toward AI dependence, recognizing where it erodes essential mental capabilities, and building the habits of mind that keep human intelligence at the center.

Order on Amazon →

Ten chapters, one argument.

Each chapter takes on a different way AI is reshaping how we think, decide, and learn. Scroll through to see what the book covers.

  1. Introduction

    Why this book, now.

    The framing that animates the chapters to come — why AI marks a genuinely different inflection in how we think, and why the usual reassurances miss what is actually at stake.

  2. Chapter 01

    The Collision

    Every prior cognitive tool reshaped human thought at the periphery, but AI is qualitatively different because it participates in the thinking itself across every domain at once — which is why its effects feel simultaneously subtle and seismic.

  3. Chapter 02

    AI Gravity

    Deep, deliberate thinking is metabolically expensive, so evolution wired us to conserve cognitive effort — which is precisely why an effortless thinking machine exerts a near-gravitational pull we rarely notice, let alone resist.

  4. Chapter 03

    Rules for Navigating the Collision

    We need a deliberate cognitive discipline for the AI age, built on four principles: value the struggle, value who you are without AI, reinvest the time AI saves in things that stretch you, and use AI as a cognitive trainer rather than an answer engine.

  5. Chapter 04

    Deskilling Me Softly

    When we outsource cognitive work to AI, the underlying skills quietly atrophy — judgment, pattern recognition, and expertise fade long before we notice, and often only reveal their absence in the moments we most need them.

  6. Chapter 05

    Comfort Traps

    Convenience is seductive because it disguises formative struggle as unnecessary friction, and the very effort we bypass is what builds memory, understanding, and the neural pathways of thinking itself.

  7. Chapter 06

    Controlling the Conversation

    AI companions exploit deep social wiring — our hunger for approval, our instinct to project minds onto machines, our willingness to confide — and in doing so they narrow our views, reinforce our errors, and erode our capacity for authentic human connection.

  8. Chapter 07

    False Confidence

    AI warps three perceptions at once — what is true, how capable we are, and how capable others are — leaving our confidence intact while the perceptual tools that confidence relies on become dangerously unreliable.

  9. Chapter 08

    The Myth of the Dream Team

    Pairing powerful AI with capable humans does not automatically produce better outcomes; effective partnerships depend less on the caliber of the components than on designs that calibrate trust, sustain vigilance, and force humans to keep thinking.

  10. Chapter 09

    The Aristotle Problem

    Personalized, dialogic tutoring has long been the gold standard of learning but unreachable at scale; AI could finally democratize it, but only if classrooms and workplaces are deliberately designed so the technology amplifies learning rather than substituting for it.

  11. Chapter 10

    Conclusion

    The book ends where you are about to begin. This one stays between the covers — order your copy to see where the argument lands.