Nobody in the room has to be wrong

The situation

A school system does its important work collectively. A panel buys the platform, a working group signs off the specification, a steering group judges the pilot, and a target tells everyone whether it worked. Each is treated as safer than one person deciding. More people means more scrutiny, and if something is wrong, somebody will have noticed.

That belief is the pitfall. I have chaired the meeting where everyone agreed and nobody believed it, and signed off a specification I would not have defended alone because seven silences looked like support. Nothing about the people in the room explains either. The room did it.

The question

What if collective work has pitfalls of its own, separate from the biases each member brings in? The Abilene paradox has no member who wanted the trip. Goodhart’s law needs nobody to cheat. A room left to itself spends its time on what everyone already knows and buries the one thing a single person brought, which Stasser and Titus showed in 1985 is usually the thing that matters. Add rank and the senior view is stated first. Add cohesion and the doubt goes into a private message afterwards.

And the damage is rarely one decision. Vaughan’s account of Challenger is a standard lowered one accepted exception at a time, until the exception was the standard the next one was judged against. Drift, rule accretion and initiative fatigue work the same way. Nobody chooses where the organisation arrives.

The move

Treat the room as something to be designed, the way a screen is. Getting the best out of it is active work, and the moves are cheap. Written positions before discussion, so nobody sees the room before committing. The senior person speaks last. The person who actually disagrees gets the floor, rather than an assigned devil’s advocate, whom a room discounts as a role. Minute the positions held before discussion as well as the conclusion.

Attach every measure to the thing it stands for. An attendance target or a platform’s minutes on task stops measuring the moment a school’s figure is reported upward. Pair the fast proxy with the slow outcome, set a floor rather than a ranking, and take any measure of learning from a low-stakes sample nobody can prepare for.

Expect drift and build for it. Date every exception and give it an expiry. Audit against the written standard, never against last month’s practice. Choose pilot sites rather than accepting volunteers, and forecast on the independent evaluation rather than the supplier’s.

What it produced

The first thing was a companion to the bias field guide: 253 collective failures in twenty families, each with a documented case, the early signs, the counter-move and what it means for a school system. Writing that last section 253 times is what turned the reading into an argument.

The second was a rule about committees. A committee is not a safeguard. It is a procedure for producing a decision, and unless the procedure is designed to surface what each member knows, what it produces is a decision with no owner and no dissent on record. When a panel is proposed as the answer to a risk, I now ask what it will do differently from a conversation, in writing, before the first meeting.

The third was some humility. A third of the collection is models rather than findings, adopted because they organise many cases well. You cannot randomise a ministry, so the field runs on inquiry reports and natural experiments, and every entry says what stands behind it on the slide rather than in a footnote.