Why do smart teams make dumb decisions?

Because intelligence is a property of individuals, and decisions made in rooms are a property of process. A team of brilliant people with a broken aggregation process will reliably lose to a team of average people with a good one. The bug is almost never “we weren’t smart enough.” It’s one of four specific failures.

The four conditions, and how meetings break them

James Surowiecki’s The Wisdom of Crowds is remembered for the fairground story — 787 villagers guessing an ox’s weight, their median landing within a pound of the truth — but the book is really about the conditions that made the trick work: diversity, independence, decentralization, and aggregation. A typical meeting violates all four at once:

• Diversity dies in hiring and promotion. Teams select people who think alike, then call their agreement “alignment.” Shared blind spots don’t cancel out; they compound.
• Independence dies the moment the senior person speaks first. Surowiecki’s information cascades are brutal: once the first two opinions point the same way, every subsequent speaker rationally weighs the crowd over their own private doubts. Smart individuals, dumb sequence.
• Decentralization dies when the room is far from the work. The people with the load-bearing details — the engineer who knows the migration is shakier than the slide says — often aren’t present, or aren’t asked.
• Aggregation dies because there isn’t any. Most meetings have no mechanism for combining views. They have a conversation, and the conclusion belongs to whoever is most confident, most senior, or most tired of the meeting. Talkativeness predicts influence; accuracy doesn’t.

What the group amplifies

Daniel Kahneman’s Thinking, Fast and Slow supplies the second half of the diagnosis: groups don’t just fail to cancel individual biases — unstructured discussion amplifies them.

• WYSIATI (what you see is all there is): the group builds a coherent story from the evidence in the room and never registers what’s missing. Coherence breeds confidence; confidence breeds consensus; nobody asks what isn’t on the slide.
• Availability cascades: the vivid anecdote told early dominates the statistics mentioned late.
• Anchoring: the first number on the whiteboard — the first revenue projection, the first deadline — drags every subsequent “independent” estimate toward it.
• Group polarization: deliberation tends to make groups more extreme and more confident than their average member started out, while accuracy stays flat or falls.

This is why post-mortems of disasters so often find that someone in the room knew. The Columbia shuttle case in Surowiecki’s book is the canonical version: the information existed at the edges of the organization; the meeting structure ensured it never aggregated into the decision.

The fixes are procedural, not motivational

The good news: every one of these failures has a cheap, boring, structural fix. None of them require better people.

1. Collect estimates in writing before anyone speaks. Private, simultaneous, specific. This single practice restores independence and creates something to aggregate. The median of the silent room is your baseline; discussion then has to argue against it, with reasons.
2. Speak in reverse order of seniority. If the leader’s view is known, the cascade has already happened.
3. Run a premortem (Kahneman, via Gary Klein): “It’s a year from now and this decision failed — each of you, write the history of the failure.” It legitimizes the doubts that politeness suppresses.
4. Hunt for the missing evidence explicitly. One standing question — “what would we expect to see if we’re wrong, and have we looked?” — is a direct counter to WYSIATI.
5. Separate information-gathering from deciding. Invite the people who carry the local details to the first; keep the decision meeting small.
6. Score dimensions independently before forming verdicts — on candidates, vendors, strategies — so one vivid strength can’t halo over everything else.

The uncomfortable summary

A unanimous, fast, confident decision by a roomful of smart people is not a good sign. It’s the signature of a cascade. The best collective decisions, Surowiecki insists, are products of disagreement and contest — and disagreement doesn’t survive in rooms without procedures to protect it. If your team’s decisions keep going wrong, stop auditing the people and start auditing the sequence in which they talk.