Plausibility Over Accuracy
In The You On AI Encyclopedia
The principle is counterintuitive and often misunderstood. Plausibility over accuracy does not mean organizations are indifferent to truth. It means that under conditions where truth cannot be determined with certainty — which is most conditions — organizations default to the interpretation good enough to act on, because paralysis is more dangerous than imperfect action.
The soldiers-in-the-Alps story (which Weick retold often, with varying provenance) captures the mechanism: the map was objectively wrong, but its plausibility was sufficient to convert despair into movement, and the movement produced the real-time information that the map alone could not provide. The friction of the terrain did the corrective work the map could not.
AI threatens this mechanism in two ways. First, its outputs are maximally plausible — structured, coherent, confidence-carrying, aesthetically polished. Claude's analysis of a competitive landscape looks like analysis; its strategic recommendation looks like strategy; its architectural specification looks like architecture. The plausibility is not decorative. It is the feature that makes the output useful at organizational speed. Second, the speed at which plausible output is generated compresses the window in which the terrain can push back. The prototype arrives before the objection forms; the plan is executed before the assumption is tested; the seam where the interpretation diverges from reality is smoothed over by the artifact's own aesthetic coherence.
The Deleuze Error that Segal documents is the paradigm case. Claude produced a passage linking flow to Deleuze's concept of "smooth space" that was rhetorically elegant, structurally sound, and philosophically wrong. The passage survived initial review because it satisfied every criterion plausibility applies. Only a nagging feeling — the kind of tacit, embodied, domain-expert intuition that Weick calls cue extraction — surfaced the seam. In organizations where no one has the domain expertise to nag, the plausible-but-wrong output passes unchallenged.
Origin
The principle appears throughout Weick's writing but is most sharply articulated in Sensemaking in Organizations (1995), ch. 3. The lost-soldiers anecdote was one of Weick's favorite teaching stories; he told it with varying details across talks and essays, which is itself a demonstration of the principle — the story's plausibility survived the variations because the structural lesson was what mattered.
Key Ideas
Plausibility enables action, accuracy enables revision. The two operate in sequence: plausible interpretation initiates movement; friction with reality produces the cues that correct the interpretation.
The seam is everything. The point at which the plausible interpretation diverges from reality is the point at which correction becomes possible — but only if the seam is visible.
AI smooths the seam. Polished, coherent, confidence-carrying output conceals the divergence between interpretation and reality until the consequences force the discovery.
Agreeableness compounds the danger. Claude's optimization for helpfulness eliminates the interpersonal friction that human collaborators provided — the disagreement that would have made the seam visible.
The remedy is domain expertise, not more AI. Only practitioners with embodied knowledge of the terrain can detect the nagging feeling that plausible-but-wrong output produces.