When it comes to the often uncomfortable discussions about the continued significance of race and ethnicity and bias in the lives of Americans, the scientific community seems to be no better at telling that complicated story than any other segment of our society. We view scientists as more rigorous, analytical and evidence-based, but we are not necessarily—especially when it comes to our own behavior.
The path from studying science to achieving the principal investigator (PI) status of a National Institutes of Health (NIH) grant is extraordinarily complicated for anyone. But in spite of many years of programs designed to increase the diversity of biomedical and behavioral scientists in the United States, the percentage of African American PIs has never exceeded 2 percent. This led to the first comprehensive analysis of the factors influencing racial and ethnic disparities in the funding of applications at NIH, and in 2011, along with five other colleagues, we published a paper in Science addressing our findings.
The paper was widely viewed as being primarily about whether or not racial bias existed in the NIH review process. While we indeed considered bias as a possible explanation for some of our findings, we emphasized that bias, if present, was not likely to be the primary driver of the differences in funding that we found. The review of grant applications is just one relatively small step in a long, complex pathway to career success as a scientist.
Unfortunately, there is such a strong impulse to reduce a complicated process—and one that most scientists know intimately—to a sound bite for or against review bias. Our latest paper, in which we were able to explain much more—but still not all—of the racial disparities using more detailed data about publication differences, will likely invoke responses along these same lines. Similarly, we anticipate that there will be the tendency among some to assume that the previous paper was “wrong” and this paper is “right.” In reality, the two papers are perfect examples of the iterative and incremental nature of almost all scientific discovery, in which we constantly try to apply new insights, methods, approaches and data to get closer to the truth about a complicated phenomenon.
Given the complexity and range of factors influencing the success of scientific careers and the evidence to date, any bias in the review process alone is not likely to be an important determinant of differences in funding rates. It is unreasonable, however, given the evidence of the pervasiveness of bias throughout numerous dimensions of life, that scientists are exempt from our tendency toward bias, and equally untenable to suggest that scientists on NIH review committees are somehow the exception and exhibit no bias in their review of grants or, more importantly, in their day-to-day activities as working scientists.
All of us see this research through the lens of our personal experiences with bias. Many of the thousands of scientists who serve on NIH review committees reject outright even the suggestion that they or their peers might be biased, despite all of the evidence of the power and pervasiveness of implicit biases. On the other hand, many scientists recall their experiences either witnessing numerous examples of bias firsthand or being the subject of bias and may think it is not at all farfetched to assume that there is substantial bias in the system of NIH peer review. And it is reasonable to conclude that our review system, grounded in human analysis and behavior, probably manifests some forms of bias. But detecting bias, especially small biases, may very well simply be beyond a definitive scientific assessment given our current methods and data.
Perhaps those of us who care deeply about the success of our nation’s biomedical scientific enterprise in an increasingly diverse country can learn from the experience of the medical and public health communities in their efforts to eliminate smoking: scientists concluded that there is no easy, simple, single causal pathway and there is no single effective, low-cost intervention. Reducing smoking rates required a wide array of research across numerous disciplines using a wide range of methods as well as multiple levels of intervention.
Likewise, addressing the disparities in career trajectories of scientists by race, ethnicity and gender will require more research across disciplines using different types of approaches and multiple levels of interventions. These range from addressing the gross disparities in the quality of K–12 educational systems by race and ethnicity, to intentionally expanding mentoring networks of underrepresented groups of scientists, to aggressively rooting out practices that result in unequal treatment of otherwise similar groups of scientists because of their race, ethnicity or gender.
Our latest analysis suggests that we should pay particular attention to ways to expand the early career networks of underrepresented scientists to help address disparities in publications, funding and careers. Mentoring programs, shown to be effective in other academic settings at reducing career disparities, are an important place to focus. We commend the NIH for investing in a range of programs—especially mentoring programs, such as the National Research Mentoring Network—that are likely to help us make progress in eliminating these career disparities.
Just as we do when addressing any other important medical, public health, or scientific problem, we don’t have to wait for the definitive answer about the precise causal pathways and subsequent randomized controlled trials in first research and then real-life settings in order to act in a way informed by what we know so far. We can move forward fully aware that future research may very well lead us to change strategies based on better evidence. These strategies and programs will likely take years to have an impact and will require more funding from NIH for additional research; even then, there will likely still be hurdles to determine their efficacy. Despite having to live with that uncertainty, scientific progress toward addressing complicated problems works by taking steps to treat the condition—we know enough to act, and so we should.