קראו את הקטע וענו על השאלות: משבר השכפול במדע

In 2015, a consortium of researchers attempted to reproduce one hundred published psychology experiments and succeeded, by the most generous count, in fewer than half. The episode gave a name to a malaise that many scientists had privately suspected: the replication crisis. Its causes are less scandalous than structural. Journals have historically preferred novel, positive findings to careful confirmations, which means a surprising result is publishable while ten failed attempts to repeat it are not. Researchers, whose careers are measured in publications, respond rationally to these incentives, sometimes massaging analyses until significance appears, a practice gentle enough to escape the label of fraud yet corrosive in aggregate. The reform movement that followed has been unusually concrete. Pre-registration requires scientists to state their hypotheses and methods before collecting data, foreclosing the quiet rewriting of questions to fit answers. Registered reports go further, with journals accepting papers based on design alone, before results exist, thereby removing the incentive to find something flashy. Early evidence suggests these instruments work: registered reports publish negative results at many times the usual rate. The deeper question is whether a culture that celebrates discovery can learn to reward verification, the slower, less glamorous work on which everything else rests.

💡 Tap an underlined word to see its meaning