In a recent column at Nature, Daniel Sarewitz worries about the effect that systematic bias is having on science: “Beware the Creeping Cracks of Bias.” According to Sarewitz, scientists can no longer point the finger at traditional causes but now should “recognize bias is an inescapable element of research.” He finds biomedical research most susceptible to bias (his claim is echoed in the comments) but implies that even physical sciences suffer from bias. Sarewitz’s bold column raises some excellent issues.
What is bias? For Sarewitz bias is over-reporting of false positive results, which result from “a powerful cultural belief … that progress in science means the continual production of positive findings.” In the biomedical realm, false positive results cannot be replicated or turn out to be invalid when applied to more complex systems. “A biased scientific result is no different from a useless one.”
According to Sarewitz, bias is not a function of scientific research so much as a characterization of the results of that research. Useful, reproducible results are not biased. Useless and irreproducible results are biased and threaten to erode public trust in science and scientists.
Such an understanding of bias is naive. Bias shapes the construction of every experiment and the interpretation of every result. Such bias is not necessarily malicious but is inescapable. Bias produces not merely systematic errors or useless results. Bias often guides the very research questions, the construction of procedures thought useful in investigating those questions, and regularly produces useful results.
Equating bias with utility ignores history and the manifold ways bias has produced useful if sometimes harmful results. For a few more infamous examples, see “What are Science’s Ugliest Experiments.”