The individual mechanism, extended
The starting point is the one this site has already made: under acute physiological arousal, deliberate reasoning narrows first, and what is left runs on pattern and instinct, the mode least equipped to catch its own mistakes. That claim is about one person. The unexamined case is a room full of people making one decision together, where the group's System 2 capacity is not simply the sum of the people in it.
Emotion moves through groups the way it moves through individuals, faster than most people account for. Sigal Barsade's foundational work on emotional contagion showed that a single person's affect, positive or negative, spreads measurably through a work group within minutes and shifts the group's cooperation and judgment, not just its mood. A room does not average its members' states. It tends to entrain toward whoever is most activated, and more so when that person also holds positional power.
What this looks like at the table
Put a board in the room. Two directors walked in already braced, one from a tense call just before, one from a personal conflict with the CEO that has nothing to do with the agenda item. Neither says a word about it. Within the first ten minutes, the room's baseline has shifted toward theirs, not because anyone agreed with their position, but because activated affect is the easiest thing in a room to transmit and the hardest to consciously discount. The vote that follows reflects the room's temperature more than the proposal's merits, and everyone in it will describe the outcome as a considered decision.
This is not a claim that boards are irrational. It is a claim that the room's accessible mode of thinking is a property of the room, set disproportionately by whoever is most dysregulated and most senior, and that this property changes before a single argument is made.
Where AI changes the failure
A dysregulated room reaches for external anchors, and a confident AI-generated brief or model output is an unusually good one. It is easier to agree with a document than to work through the room's own unspoken tension, and a document that sounds authoritative gives everyone a face-saving way to end a disagreement nobody wants to have out loud. The output gets adopted as arbitration, not because it settled the substantive question, but because it let the room stop feeling what it was feeling. This is the group-level version of the same failure named elsewhere on this site: the tool arrives exactly when scrutiny is least available, and its fluency gets mistaken for having resolved something.
The practical move
The individual version of this is a state check before a high-stakes call. The group version needs a second move, because a room cannot self-regulate the way one person can. Whoever is chairing has to be willing to name the room's temperature out loud before opening a vote, not after a bad one. That is an uncomfortable thing to say in a boardroom, and it is also the only intervention that works at the level where the actual failure is happening. Reading the agenda item more carefully does not fix a problem that was never in the agenda item.
The same test applies here as everywhere else on this site. A board that gets better, over time, at noticing its own temperature before it decides is building a real capability. A board that gets more dependent on whoever brings the most confident document is not, whatever the document says.