One of the most surprising discoveries in twentieth-century psychology is that a simple regression equation usually produces better results than a human expert, a fact that was first discussed by psychologist Paul Meehl .
Regression analysis is a statistical process that allows us to analyze the relationship that exists between two or more variables, one of them being dependent on the rest of the variables that we are using in our mathematical calculation. In other words, a regressive analysis makes it possible to understand how independent variables directly affect another variable that depends on them.
Thanks to this powerful statistical tool, it is possible to predict how long a cancer patient will survive, or if a defendant will abscond while on bail. An expert can also try to predict these events, but an equation is simply more accurate, as Steven Pinker points out in his book Rationality :
We have a set of predictors: a checklist for symptoms, a set of demographic characteristics, a count of past behaviors, a college transcript or test results, or whatever may be relevant to the prediction challenge. We show the data to an expert (a psychiatrist, a judge, an investment analyst, etc.) and, at the same time, we feed it into a standard regression analysis to obtain the prediction equation. (…) The winner is, almost always, the equation.
The most surprising thing is that if the expert also uses the equation to express his judgment, he also tends to get worse results than the equation alone . The reason is that experts are too quick to see extenuating circumstances or striking details that, in their opinion, make the formula in question inapplicable.
This happens because the algorithm already takes mitigating circumstances into account. Not to mention that some of the predictors most trusted by human experts, such as face-to-face interviews, are perfectly useless .
Despite all this, there are people who trust people more. In the professionals. Even in human judgment. And, on the other hand, he considers the guidelines or algorithms to be cold, rigid and allergic to empathy. All these romantic ideas can be easily dismantled if we take a look at how judges, doctors or pilots are glaringly wrong as soon as they stop mechanizing their judgments, diagnoses and decisions :