Noise in judgments and bias in the decisions of people and neural networks

When software developers were asked to estimate the time it would take to complete a task for the same specification on two different workdays, the hours they predicted differed by an average of 71%.

Two equally experienced doctors can give two different diagnoses to the same patient, two competent personnel officers can invite two completely different candidates for the same position, two experts can give noticeably different forecasts about sales figures in the next quarter, two judges can assign different sentences to defendants. And even the same person on different days of the week can solve the same problem in a completely different way.

The reason is “noise” in judgments, that is, dispersion in decisions that, by all rules, should be the same.

Daniel Kahneman, Nobel laureate 2002, spoke about this very clearly.

Kahneman is one of the main modern economists, although he himself actively denies this. He destroyed a bunch of theoretical ideas about how people make decisions and created behavioral economics.

And even more. It turned out that the principles that Kahneman described in his books and scientific works are valid not only for human thinking, but also for artificial intelligence. And they are already being widely used in training neural networks.

Today we’ll talk about rationality, errors in judgment, ways to reduce “noise” and the “unicorn” of economics – Daniel Kahneman.