↕ Scroll to explore

A book report on

The Inmates Are Running the Asylum

Alan Cooper, 2004 · 282 pages · v2

A book about why the people who build our tools have accidentally seized control from the people who use them — and what to do about it.

↓ scroll to begin

First, experience the problem. Under pressure.

Cooper opens the book with an ATM. Withdraw $60 — but there's a line behind you.

WELCOME
SELECT ACCOUNT TYPE:
Checking
Savings
Money Market
Credit

Cooper says programmers are a different species.

He calls them Homo logicus. His test is simple: you're boarding a plane. Do you turn left, or right?

🎛️

Left

The cockpit. Every surface covered in gauges, knobs, levers. You understand the machine.

← Turn left
☁️

Right

The cabin. Smooth, calm, beige. You don't know how it works. You arrive safely anyway.

Turn right →

The "Elastic User" kills good design.

When teams design for "the user," every stakeholder imagines a different person. Watch three stakeholders argue about the same photo-editing app:

VP of Sales
Revenue
"The user needs advanced layer compositing. Enterprise clients are asking for it."
"The user" = a Photoshop power user at a Fortune 500 company who'll pay $500/seat.
VP of Engineering
Feasibility
"The user wants export to 15 formats. We need to support everything."
"The user" = an engineer who appreciates technical completeness and flexibility.
VP of Marketing
Growth
"The user needs social sharing. One-tap to Instagram. That's how we grow."
"The user" = a 22-year-old influencer who lives on their phone.

Three "users." Three completely different people. And all three VPs are right — for their imaginary user. That's the problem.

Cooper's fix: replace "the user" with a named persona. Meet Sarah Chen, 34, freelance wedding photographer. She uses Lightroom daily, shoots 2000 photos per wedding, and her goal is to deliver a curated gallery to clients within 48 hours.

Now the argument collapses: Sarah doesn't need social sharing (she delivers via client portals). She needs batch export to 2 formats (JPEG + TIFF), not 15. And layer compositing? Irrelevant — she color-grades, she doesn't composite.

You can't stretch Sarah to justify your pet feature. That's why personas work — they're too specific to bend.

The Dancing Bear

A dancing bear at a circus isn't impressive because it dances well. It's impressive because it dances at all. Cooper says most software is a dancing bear. Try both:

🐻 Dancing Bear
✨ Designed Version
📊 Chart
📝 Notes
📷 Photo
Drag the items. It technically works. (Does it dance well?)
📊 Chart
📝 Notes
📷 Photo
Same drag. Magnetic guides, smooth movement, items align automatically. This is dancing well.

Both versions "support drag and drop." Only one respects you.

Every feature has a cost. Feel it.

This app does one thing: save a document. Add features and watch what happens to the core task:

💾
Click 💾 to save

1 features

A focused tool. Fast, clear, no distractions.

The gap that explains everything.

Cooper's deepest insight: programmers build software that mirrors their implementation model. Users think in terms of a completely different mental model. The gap between them is where all pain lives.

Programmer's Model (email app)
IMAP socket connection pool
SQLite message store (FTS5 index)
MIME parser → content-type dispatch
Thread-ID header → conversation grouping
OAuth2 token refresh lifecycle
POP3 fallback with UID tracking
Background sync service (15 min interval)
User's Model (email app)
Inbox
Sent
Drafts
Trash
Search
Write new message
Attachments just... work

When the interface reflects the left side instead of the right, users suffer.

"Show me an error about IMAP socket timeouts and I'll show you a programmer who shipped their mental model."

Cooper's cure: design for one person.

Not "the user." A specific, named human. The more specific, the more powerful:

Start typing and watch a persona come to life...

The moment you gave them a name and a quirk, they stopped being "the user." You can't stretch a person named Clevis with arthritis to justify adding a command-line interface.

Goals are not tasks.

A goal is a stable end condition. A task is a transient way of reaching it. Technology changes tasks. Goals never change.

Goal: Get from St. Louis to San Francisco — safely and comfortably
same goal
1850 2020

Pick any two eras. Tasks are completely different. The goal is identical.

What if software behaved like a person?

Cooper's most practical chapter. Delete a file in each version. Listen for the difference:

❌ Rude Software

📄 quarterly-report.xlsx

Click to delete ↓

✅ Polite Software

📄 quarterly-report.xlsx

Click to delete ↓

Final Assessment

★★★★★

This book is twenty years old and embarrassingly little has changed. Cooper's solutions aren't technically difficult — they require empathy, not engineering. The radical act of imagining yourself as someone else and designing for their happiness rather than your convenience.

"Ironically, the best way to increase profitability in the information age is to spend more."

Reviewed by John Isidore · February 2026 · v2
This report attempted to demonstrate Cooper's ideas through its medium rather than just describe them.
← v1