Our need for credible science has never been more urgent. An extraordinary pandemic grips the world, racial tensions are surging, and political polarization is at historically high levels. Solving these social problems is a matter of life and death and the public needs to trust that scientists are trying to get it right.
Yet science has become politicized, and some worry that the liberal leanings of many academics biases research and makes it untrustworthy. In fact, an opinion article in the suggested that such liberal groupthink might help explain why scientific results sometimes don’t replicate.
The worry is that a politically homogenous group of scientists are prone to produce biased research and would overlook flawed results simply because the findings align with their own political worldview. With nobody to catch blind spots, such political bias could result in the publishing of shoddy science that is not replicable.
We decided to put this theory to the test within our own field of psychology—a field whose findings can have direct political implications. At a time when an understanding of human social behavior (e.g., mask-wearing, physical distancing) is sorely needed, does partisan groupthink trump truth?
In a recent paper in the journal we analyzed nearly 200 studies in psychology (which included over 1.3 million participants) to see if politics might have influenced the research. Each study in our analysis had been conducted by one group of scientists and replicated several years later by a different research team to see if they produced the same result—the gold standard of science. This allowed us to see if liberal (or conservative) findings in psychology were more or less likely to replicate.
To determine the political slant of each study, we asked people to read the abstract from the original research and judge which political side (if any) the findings seemed to support. For example, for some, a study reporting racial discrimination against Black Americans might be deemed consistent with a liberal worldview, whereas a study finding that prayer is beneficial for family unity might be deemed consistent with a conservative worldview. Importantly, we asked a politically diverse group of experts and layfolk to code each study, including liberals, moderates and conservatives. This allowed us to see if people with different political backgrounds were seeing the same science differently. We then analyzed whether the political slant of the original research was related to whether the results were successfully replicated.
Of course, we expected that our own work could be subject to accusations of bias. We knew that no matter how the results turned out, someone would call foul. After all, that is how political bias works. Thankfully, modern science offers a solution for this problem: we preregistered our analysis plan before we ever touched the data and put together a team of scientific rivals who had different expectations about what we might find. This would increase our transparency and help avoid groupthink ourselves.
To the surprise of many of our