The Four Blindnesses: Four kinds of institutional blindness, and the remedy for each

I have reorganized to fix a problem a reorganization could never touch. Something kept getting missed in one of my operations, so I redrew the boxes, changed who reported to whom, and enjoyed the pleasant sensation of doing something. Months and real money later, the same thing got missed again. The signal had been arriving the entire time. It was getting buried because the person who saw it was quietly rewarded for not raising it. No new box fixes that. I had built a sensor plan for an incentive problem.

This book sorts institutional blindness into four kinds, because the cure depends on which one you have. A sensor blindness has no instrument to catch the signal, so the data never arrives. An incentive blindness lets the signal arrive and then smothers it, the case Upton Sinclair caught in 1934 when he wrote that a man will not understand a thing his salary depends on his not understanding. An ideological blindness rules a possibility out before any signal is weighed, which is how James Scott's high-modernist states failed and how Diane Vaughan's normalization of deviance walked NASA into the Challenger launch in 1986. A temporal blindness is a mismatch between the clock the institution reports on and the speed of the thing it has to catch. Goodhart's law, the McNamara fallacy, and hyperbolic discounting each describe a different one of these.

Name the kind before you prescribe. Reach for structure by reflex and you get the flailing reorganization: motion that feels like progress and moves nothing.

This is a short book about the four institutional blindnesses, sensor, incentive, ideological, and temporal, and about the principal-agent problem, normalization of deviance, groupthink, Goodhart's law, and the McNamara fallacy, each diagnosed before any remedy is prescribed.

This book is written from an Indonesian operator's context, but the reflex to reorganize before diagnosing turns up in institutions of every size and market.