The most expensive thing that happens in a leadership meeting is invisible. The CEO opens with a view, or with a slide that only makes sense if the answer is yes. Four smart people then adjust what they were going to say by a few degrees each, and none of them notices they have done it. The meeting ends with a decision that looks unanimous and was actually made by one person, with four people's silence attached to it.
I did this for years as a president. I thought I was being efficient by framing the question. What I was doing was answering it.
The order of speaking decides what the room learns
In 2011, Jan Lorenz and colleagues at ETH Zurich ran an experiment that should be required reading for anyone who chairs a meeting. They asked people factual questions with numerical answers, then let some of them see what others had guessed before revising. On the first round, the group's average was close to the truth, the classic wisdom-of-crowds result that James Surowiecki popularized in 2004. Once people could see each other's answers, the guesses converged. The spread narrowed. The group became more confident. And the average moved away from the truth. Social influence made the group agree, and agreement turned out to be a poor substitute for accuracy.
The finding is about information rather than intelligence. Each person's first guess carries a little private knowledge and a little error. Averaged, the errors cancel and the knowledge adds up. Let people see each other first and the private knowledge gets overwritten by whoever spoke earliest or loudest.
Stasser and Titus had found the same thing from the other direction in 1985. They gave groups a hiring decision where some facts were known to everyone and some to only one member. The groups talked mostly about the shared facts. The unshared ones, the ones that would have changed the answer, mostly never came up. A meeting is a machine for repeating what everyone already knows.
Why the assigned devil's advocate does not fix it
The usual remedy is to nominate somebody to argue against. Charlan Nemeth, who spent a career studying dissent at Berkeley, tested this in 2001. Groups with a genuine dissenter, someone who actually held the minority view, produced more solutions and better ones. Groups with an assigned devil's advocate, someone playing the part, did not. Worse, people in those groups became more confident in the original view, because they felt they had heard the other side.
Performed disagreement is a ritual. Real disagreement is information. You cannot assign the second kind; you can only make it safe to bring.
What independent input looks like in practice
Kahneman, Sibony and Sunstein call the fix decision hygiene, and the core of it is simple: collect each person's judgment before any of them has heard another's. They describe it as running a statistical jury rather than a deliberating one. The jury still deliberates. It just deliberates after the votes are in, so the deliberation has something to work with other than the first confident voice.
For a company of forty people the practical version fits in an afternoon and adds no meeting.
Write the decision down as a handful of specific claims: what would have to be true for the chosen path to work. Not "should we do this?" but "customers in the mid-market will pay 15 percent more without churning" and "we can hire the two engineers by March".
Send each person the claims and ask for their read on each one, alone, with a deadline. Make "I don't know" a complete answer, so nobody has to manufacture an opinion. A silence is fine too, as long as you keep track of who was asked and who answered, because those are different facts.
Hold the answers until everyone's is in. Then look at the spread. Where four people agree, move on. Where they split, that is the conversation, and it is a conversation about a claim rather than about the CEO's judgment.
Only then bring the room together, and start from the disagreement. The meeting gets shorter, because it skips everything the answers already settled.
Where Midfire fits
This is the reason Midfire asks colleagues one claim at a time and keeps every answer sealed until the person's own is in. The CEO's view is not visible to the people asked, and "I don't know" is one of the answers on the button. The product does not make anyone braver; it removes the moment where bravery would have been required.
The uncomfortable part
Independent input will show you, in writing, how often your team's honest read differs from what they say in the room. The first time I saw the spread on a decision I had thought was obvious, I was annoyed for a day and grateful for a year. The disagreement was there all along. The order of speaking had hidden it.