Une étude effrayante – why we pretend all opinions are equal

Je me demande si nos ancêtres étaient aussi stupide que nous, et quelque part j’ai un gros doute. Cette étude montre pourquoi les choses vont de pire en pire. 

« The new study, just published in the Proceedings of the National Academy of Sciences, is by Ali Mahmoodi of the University of Tehran and a long list of colleagues from universities in the UK, Germany, China, Denmark, and the United States. And no wonder: The research was transnational, and the same experiment — with the same basic results — was carried out across cultures in China, Denmark, and Iran.

 In the experiment, two separate people view two successive images, which are almost exactly the same, but not quite. In one of the images, there is an “oddball target” that looks slightly different. The images flash by very fast, and the two individuals have to decide which one, the first or the second, contained the target.

Sounds simple enough — but the two individuals didn’t merely have to identify the target. They also had to agree. Each member of the pair — the scientists wonkily call it a “dyad” — separately indicated which of the images contained the target, and how confident they were about that. Then, if there was a disagreement, one individual was chosen at random to decide what the right answer was – and thus, who was right and who was wrong. And then, both individuals learned the truth about whether their group decision had been the correct one or not.

This went on for 256 intervals, so the two individuals got to know each other quite well — and to know one another’s accuracy and skill quite well. Thus, if one member of the group was better than the other, both would pretty clearly notice. And a rational decision, you might think, would be for the less accurate group member to begin to favor the views of the more accurate one — and for the accurate one to favor his or her own assessments.

But that’s not what happened. Instead, report the study authors, “the worse members of each dyad underweighted their partner’s opinion (i.e., assigned less weight to their partner’s opinion than recommended by the optimal model), whereas the better members of each dyad overweighted their partner’s opinion.” Or to put it more bluntly, individuals tended to act “as if they were as good or as bad as their partner” — even when they quite obviously weren’t.

The researchers tried several variations on the experiment, and this “equality bias” didn’t go away. In one case, a “running score” reminded both members of the pair who was faring better (and who worse) at identifying the target — just in case it wasn’t obvious enough already. In another case, the task became much more difficult for one group member than the other, leading to a bigger gap in scores — accentuating differences in performance. And finally, in a third variant, actual money was offered for getting it right.

None of this did away with the “equality bias”.  »

Ici