If You Reward Big Promises, You're Training People to Lie

Confidence: hypothesis. The failures below are drawn from published research and widely reported industry patterns, not firsthand deployment. The pattern at the end is an argument, not a shipped system. Argue with it.

Here is a quiet way to corrupt your own data. Build a scoring system that rewards big numbers, and let the people you are scoring know that bigger is better. They will give you bigger. Not because they are dishonest, but because you told them what wins, and they would like to win. Within a cycle or two, your inputs are inflated, your scores are built on the inflation, and the honest people are losing to the optimistic ones. You did that. The system did exactly what you designed it to do.

Choosing clinical trial sites shows this in slow motion.

The clinical why

A trial picks sites partly on their promised enrollment. A site that says it can enroll twenty patients looks better than one that says eight. So sites learn to say twenty. The one that answers honestly, “eight, realistically,” loses the contract to the one that answers “twenty” and then delivers four. The honest site got punished for being honest, and everyone watching learned the lesson.