Cognitive psychologist Philip Johnson-Laird spent decades arguing that humans don't think in formal logic — we think in mental models, and crucially, we run those models like simulations, checking them against imagined scenarios rather than abstract rules. The trouble is: simulations can only test what you thought to build into them. A model that was constructed during calm, hierarchical conditions will simulate calm, hierarchical conditions. It will return confident outputs when the conditions it was designed for no longer exist. Johnson-Laird called this the 'verification problem' — we tend to search for scenarios that confirm a model runs smoothly, not scenarios that break it. The leadership implication cuts deep. The mental models most leaders trust are the ones that have been rewarded before, which means they've been tested in the situations that already went well. What they've almost never been systematically tested against is the scenario that feels structurally familiar but is actually a different category of problem disguised by surface similarity. The discipline, then, is not to accumulate more models — it's to deliberately construct adversarial simulations for your best ones. Pick your most relied-upon mental model for reading people, or diagnosing organizational problems, and build the specific counter-scenario designed to break it. Not to stress-test it in general, but to find the exact conditions under which it produces confident, plausible, wrong outputs.

Which mental model do you apply most automatically — and what type of situation has it never actually been tested against?

Drawing from Cognitive psychology / philosophy of mind — Philip Johnson-Laird