Humans are remarkably good at self-deception. But growing concern about reproducibility is driving many researchers to seek ways to fight their own worst instincts.
In 2013, five years after he co-authored a paper showing that Democratic candidates in the United States could get more votes by moving slightly to the right on economic policy1, Andrew Gelman, a statistician at Columbia University in New York City, was chagrined to learn of an error in the data analysis. In trying to replicate the work, an undergraduate student named Yang Yang Hu had discovered that Gelman had got the sign wrong on one of the variables.
Gelman immediately published a three-sentence correction, declaring that everything in the paper’s crucial section should be considered wrong until proved otherwise.
Reflecting today on how it happened, Gelman traces his error back to the natural fallibility of the human brain: “The results seemed perfectly reasonable,” he says. “Lots of times with these kinds of coding errors you get results that are just ridiculous. So you know something’s got to be wrong and you go back and search until you find the problem. If nothing seems wrong, it’s easier to miss it.”
This is the big problem in science that no one is talking about: even an honest person is a master of self-deception. Our brains evolved long ago on the African savannah, where jumping to plausible conclusions about the location of ripe fruit or the presence of a predator was a matter of survival. But a smart strategy for evading lions does not necessarily translate well to a modern laboratory, where tenure may be riding on the analysis of terabytes of multidimensional data. In today’s environment, our talent for jumping to conclusions makes it all too easy to find false patterns in randomness, to ignore alternative explanations for a result or to accept ‘reasonable’ outcomes without question — that is, to ceaselessly lead ourselves astray without realizing it.
Failure to understand our own biases has helped to create a crisis of confidence about the reproducibility of published results, says statistician John Ioannidis, co-director of the Meta-Research Innovation Center at Stanford University in Palo Alto, California. The issue goes well beyond cases of fraud. Earlier this year, a large project that attempted to replicate 100 psychology studies managed to reproduce only slightly more than one-third2. In 2012, researchers at biotechnology firm Amgen in Thousand Oaks, California, reported that they could replicate only 6 out of 53 landmark studies in oncology and haematology3. And in 2009, Ioannidis and his colleagues described how they had been able to fully reproduce only 2 out of 18 microarray-based gene-expression studies4.