Here is an uncomfortable question to ask about the last big decision your team debated in a meeting: when did you decide? Not when the group decided — when you did, privately, in your own head. If you're honest, the answer is probably "before the meeting started." You walked in leaning one way. You listened the way a lawyer listens to opposing counsel — scanning for weaknesses, not for truth. And you walked out believing exactly what you already believed, except now with witnesses.
Most meetings don't make decisions. They ratify them. The discussion feels like deliberation, the whiteboard fills up, someone says "let's pressure-test this" — and underneath it all, everyone in the room is quietly running the same program: find support for what I already think. That program has a name, it's one of the most documented phenomena in cognitive psychology, and meetings are almost perfectly designed to feed it.
The bias that feels exactly like thinking
Confirmation bias is the tendency to search for, interpret, and remember information in ways that support what we already believe. The psychologist Raymond Nickerson, in a landmark review, called it "a ubiquitous phenomenon in many guises" — and the guises are the problem. It doesn't feel like bias from the inside. It feels like diligence.
The classic demonstration comes from Peter Wason in the 1960s. He gave people the number sequence 2-4-6 and asked them to figure out the rule behind it by proposing their own sequences, which he would confirm or reject. Most people guessed the rule was "even numbers ascending by two" and then tested it with sequences like 8-10-12 and 20-22-24. Confirmations piled up; confidence soared. Almost no one tried a sequence designed to fail — like 1-2-3 or 10-9-8. The actual rule was simply "any ascending numbers," and the only way to discover it was to test sequences you expected to be wrong.
Researchers later called this the positive test strategy: we check our beliefs by looking for hits, not misses. It isn't laziness or dishonesty. It's the default setting of human hypothesis-testing. And every test that comes back positive feels like rigor.
How a meeting becomes a ratification machine
Now put eight people running that strategy in a conference room. The distortion starts before anyone speaks, in the agenda itself. "Discuss rollout plan for the new pricing" is not a question — it's a conclusion wearing a question's clothes. The meeting is framed to plan the thing, not to ask whether the thing should exist.
Then the questions begin, and they lean the way questions lean. Mark Snyder and William Swann showed in the 1970s that when people were told an interviewee was probably an extravert, they asked questions that could only surface extraverted answers — "What would you do to liven up a party?" The hypothesis shaped the evidence-gathering, and the evidence obligingly confirmed the hypothesis. Meetings do this constantly. "How would this scale?" is a different question from "Would this scale?" The first assumes success and asks for elaboration. Rooms that favor an option ask it questions it can only pass.
Even genuinely mixed evidence doesn't rescue the room. In a famous 1979 study, Charles Lord, Lee Ross, and Mark Lepper had supporters and opponents of capital punishment read the same set of studies — some supporting deterrence, some undermining it. Both sides rated the studies that agreed with them as more rigorous, picked apart the ones that didn't, and left more convinced than they arrived. The researchers called it biased assimilation. So "let's just look at the data together" isn't the safeguard it sounds like: the same slide deck can harden two opposing camps at once, each side scoring its own points.
Being smart doesn't save you
Here is the part that stings. Research by Keith Stanovich and Richard West on what they call myside bias found that the tendency to evaluate evidence in favor of your own position shows little to no correlation with intelligence. Smart people are not less biased — they are better armed. Greater cognitive ability means a faster, more fluent generation of reasons... for your side. The senior engineer who can construct a five-step argument in real time is not necessarily closer to the truth; they are simply harder to stop.
The psychologist Thomas Gilovich captured the underlying asymmetry neatly: for conclusions we want to believe, we ask "Can I believe this?" — and almost anything clears that bar. For conclusions we'd rather reject, we ask "Must I believe this?" — and almost nothing does. Both questions get answered honestly. The rigging happened in the choice of question.
What actually changes minds in a room
The standard fix — appointing a devil's advocate — mostly doesn't work. Charlan Nemeth's research on dissent found that role-played, assigned opposition stimulates far less genuine rethinking than authentic dissent from someone who actually disagrees. Worse, it can backfire: once the group has "heard the other side" from a designated actor, members feel inoculated and grow more confident. The ritual of doubt substitutes for the experience of it.
What does work, according to the evidence, is changing the task rather than the casting. In a follow-up to the biased assimilation study, Lord, Lepper, and Elizabeth Preston found that telling people to "be fair and unbiased" changed nothing — but instructing them to consider the opposite ("ask yourself how you'd evaluate this study if it had come out the other way") measurably reduced the distortion. Fairness is an intention; considering the opposite is a procedure.
The other procedure with real teeth is Gary Klein's premortem. Instead of asking "what could go wrong?" — a question confident rooms answer with token risks — the premortem announces: "It's a year from now. This project failed badly. Write down why." That temporal trick, sometimes called prospective hindsight, does two things. It converts imagination from advocacy into explanation, and it makes doubt a completed assignment rather than a social risk. The skeptic no longer has to volunteer as the room's problem; everyone has been ordered to be one for ten minutes.
Your next moves
- Write down your prior before the meeting, with an exit condition. One sentence: "I currently lean toward X. The evidence that would change my mind is ___." If you can't fill in the blank, you don't hold a position — you hold an allegiance, and the meeting will only feed it.
- Rewrite one agenda item from a confirming question to a disconfirming one. "Discuss rollout of the new pricing" becomes "What evidence would make us kill the new pricing?" Presume nothing survives the meeting by default.
- Run a ten-minute premortem on your biggest live decision. Declare the project dead a year from now, have everyone silently write the causes of death, then read them aloud before any discussion. Silence first — otherwise the loudest optimist frames the exercise.
- Assign evidence, not roles. Skip the devil's advocate. Instead, ask one person to arrive with the three strongest verifiable facts against the favored option — real data they had to go find, not objections improvised in the room.
- Keep a private tally in your next meeting. Two columns: questions that seek support versus questions that seek problems. Most people who try this are startled by the ratio — and once you see it, you can be the person who asks the other kind.
Seeing the ratification from outside
The cruelest property of confirmation bias is that it is invisible from inside the meeting — the room that most needs to notice it is the room least equipped to. It only becomes legible afterward, in the record: what was actually asked, who actually dissented, whether the "decision" was ever genuinely open. That's the idea behind meetingmortem — a simple way to run an honest autopsy on your meetings, so you can see which ones deliberated, which ones merely ratified, and what to change before the next one. The exercises above cost nothing and work today; if you want help making the pattern visible week after week, meetingmortem is there when you're ready.