For fake checks in the double-digit dollar range, fraudsters in the USA have to sit in prison for up to 15 years. Or only 30 days, depending on the judge. And depending on the person in charge, insurance premiums for comparable cases can differ by up to 55 percent. Similar phenomena can be found in asylum applications, economic forecasts, patent approvals, lending and medical diagnoses.
This spread can partly be explained by “bias” (prejudice), for example in the case of judges who generally judge particularly strictly. But at least as great a source of error is “noise”, write Daniel Kahneman, Olivier Sibony and Cass R. Sunstein – So the random fluctuation of one and the same decision maker depending on the daily form.
“Organizations around the world consider bias to be something to be resolved upon. You’re right. But they don’t see the same grievance in noise. But they should, ”write the authors. In many areas of life they account for a “downright scandalous level of noise”. The problem with this is that noise is more difficult to detect than bias. “It only becomes visible when we consider a set of similar judgments.”
This article is from issue 5/2021 of the Technology Review. The magazine will be available from July 8th, 2021 in stores and directly in the heise shop. Highlights from the magazine:
Kahnemann and his co-authors see the self-love of many executives, especially the most high-ranking and most experienced, as the central cause of noise. They still trust their “intuition” or their “gut feeling” even when missing data make a reliable prognosis objectively impossible. In doing so, according to the authors, they “tacitly deny the existence of uncertainty”.
Kahneman, winner of the 2002 Nobel Prize in Economics, analyzed in his bestseller Fast Thinking, Slow Thinking ten years ago how error-prone this intuitive thinking style is. And so much of the advice for better “decision hygiene” is to replace spontaneous judgments with rules and algorithms. School essays, for example, could be judged more fairly if they were ranked: essay A is better than essay B, but not as good as essay C. Teamwork. In the end, all you have to do is add up the points.
“Decision-making is often characterized by a downright scandalous level of noise.”
Such “mechanical” decisions, however, often meet with resistance because people feel offended. The authors counter this: “Algorithms are less prone to error than human judgment. If we nevertheless have an intuitive preference for people, we should carefully review our intuitive preferences. ”
The fight against noise is always associated with costs, for example in the form of more complex processes. “Decision-making hygiene is invaluable, but at the same time an ungrateful effort,” writes the trio of authors. Because, like washing your hands, you never know exactly which pathogens or errors you have protected yourself against: “Noise is an invisible enemy against whom you can only achieve an invisible victory.”