Update (December 21): The authors retracted the paper today after three experts reviewed the study post-publication and agreed with criticisms about the use of coauthorship as a proxy for mentorship. “Although we believe that all the key findings of the paper with regards to co-authorship between junior and senior researchers are still valid, given the issues identified by reviewers about the validation of key measures, we have concluded that the most appropriate course of action is to retract the Article,” the authors write in the retraction notice, adding that they regret the pain the paper caused among members of the scientific community.
A study published in Nature Communications November 17 has caused an eruption of outrage among scientists. The study of 3 million mentor-protégé pairs in STEM found that women trainees who coauthored papers with senior women scientists received fewer citations after they became principal investigators than did women trainees who coauthored with senior men. That gendered pattern is consistent with previous research documenting citation biases in science. But what caused a firehose of criticism was the conclusion of the paper, in which the authors suggest that the solution to this disparity is that women should avoid female mentors, thus perpetuating an existing bias.
“Instead of coming to the conclusion that their data shows the system is biased, they come to the conclusion that women shouldn’t be mentors, which is blaming individuals as opposed to blaming a broken system. That’s why I think it’s so harmful,” says Rebecca Barnes, a geoscientist at Colorado College and co-PI of a National Science Foundation grant that supports a mentoring program for women in her field. Barnes and her colleagues published a 2020 study showing that exposure to women role models increases the likelihood that women trainees will be retained in geosciences.
For the Nature Communications study, researchers analyzed a dataset of 215 million scientists and 222 million papers. They used a computer program to infer the gender of scientists based on their names and designated them as junior or senior scientists (junior scientists were defined as scientists in the first seven years since their first academic publication; senior scientists were beyond that). Mentorship was defined by authorship patterns: when a junior scientist coauthored a paper with a senior scientist, the senior scientist was considered a mentor. In this way, the researchers identified 3 million unique mentor-protégé pairs spanning more than a century.
The idea that the goal of mentorship is to maximize the impact of citations later in a protégé’s career is a very shaky target.—Roberta Sinatra, IT University of Copenhagen
To probe whether these authorship patterns were a good measure of mentorship, the researchers also sent a survey to 2,000 scientists in the dataset, 167 of whom responded, that confirmed these authorship patterns involved some form of mentorship.
The researchers then measured the standing of mentors by their average number of citations per year before their first publication with the protégé. Highly cited mentors were called “big shot” mentors by the paper’s authors. The researchers also calculated the degree to which the mentor was integrated within networks of scientific collaborations (that is, how much of a “hub” each mentor is in a collaboration network). The mentorship outcome was measured as the number of citations of the protégé’s publications published beyond the first seven years of their career and without their mentor as a coauthor.
The study’s authors found that “big shot” mentors (and “hub” mentors to a lesser degree) were associated with a greater number of citations of their protégés post-independence. The researchers also found that for all trainees, “having more female mentors is associated with a decrease in the mentorship outcome, and this decrease can reach as high as 35%,” they write in their report.
Next, the researchers examined how the mentorship relationship benefited the mentors. “The results suggest that, by mentoring female instead of male protégés, the female mentors compromise their gain from mentorship, and suffer on average a loss of 18% in citations on their mentored papers,” the researchers write. Male mentors did not see an effect on their citations related to the gender of their protégés.
Scientists have heavily criticized the study on methodological grounds as well as on the authors’ interpretations and policy recommendations. Methodologically, “the study conflates authorship with mentorship,” explains Regina Baucom, who studies plant genetics at the University of Michigan.
“Lots of coauthors do zero mentoring,” Brian Uzzi, who researches factors that affect women’s success in science and business at the Kellogg School of Management at Northwestern University, writes in an email.
Uzzi and colleagues published a paper in PNAS earlier this year about mentorship and protégé success that found no differences in the success of women scientists who had male versus female mentors. In that study, the researchers used data on mentors and protégés from the ProQuest PhD Dissertation & Thesis databank, an official record of advisor-student relationships taken from doctoral theses. They measured several aspects of protégé success, including winning a scientific prize, election to the National Academy of Sciences, and a high citation impact score. “Our paper contradicts their paper’s finding and recommendations that women scientists do better when mentored by men when a woman’s actual and known mentors are identified,” Uzzi writes in an email to The Scientist.
The study’s use of algorithms to identify gender based on names is also problematic because it can lead to mistakes, Baucom and others have pointed out.
Roberta Sinatra, a scholar at the IT University of Copenhagen, uses big data to study publication, citation, mentorship, and success in science. She says that the methods used in the study, although flawed, are fairly standard in this field. The lack of gender information is a limitation of the dataset, but the algorithms are about 80 percent accurate, and questions about gender bias couldn’t begin to be addressed without this approach, she says. She commends the researchers for taking the extra step to do a survey to confirm that the authorship patterns do in fact reflect some measure of mentorship, but suggests that “chaperone,” rather than “mentor,” could be a more apt description of that relationship. Sinatra herself used that term in a 2018 study on publishing in high impact journals because, she says, the authorship patterns don’t reveal anything about the quality of mentorship.
Citations as success
“Overall, it’s a very thorough and comprehensive analysis,” Sinatra says, adding that she felt some of the methodological criticisms were unfair because the approaches are fairly standard in the field known as the science of science. But she agrees with critics who say the researchers should have been more careful in their interpretation of the results. “The interpretation starts from the idea that scientists have to achieve a higher number of citations. I think that is not a good starting point,” she says. “The idea that the goal of mentorship is to maximize the impact of citations later in a protégé’s career is a very shaky target.”
Sinatra suspects that the fact that female protégés mentored by women experienced less citation impact later is likely due to an “amplification of bias,” though more analysis would be needed to confirm that. But she emphasizes that good mentoring isn’t just about citations; for example, the retention of women and underrepresented minorities is an important aspect of good mentorship that is not captured in the analysis.
The researchers did not control for well-documented structural issues and biases that women in science face with respect to hiring, publication, citation, funding, promotion, and more, says Daniel Acuña, who studies decisionmaking related to scientific publication, collaboration, and funding at Syracuse University, and who was one of the paper’s peer reviewers.
“To me, the paper seems irresponsible to the point of malfeasance,” Baucom says. “The researchers didn’t have a control and a treatment group, so they can’t assess the reasons why women might not be as highly cited. To take that and say to young women, ‘You shouldn’t go work with a woman mentor’—that, to me, is malfeasance,” she says. She adds that in this moment in particular, when the pandemic has amplified the structural and societal obstacles that women in science are facing, publishing this paper was a mistake.
In response to an interview request, the authors of the paper, Bedoor AlShebli, Kinga Makovi, and Talal Rahwan of the Computational Social Science Lab at New York University Abu Dhabi, write in an emailed statement: “In our paper, we highlight that the elevation of women in science depends on the achievement of at least two objectives: retaining women in scientific careers—for which female mentors are indispensable, as explicitly mentioned in our paper—and maximizing women’s long-term impact in the academy. As we conclude: ‘the goal of gender equity in science, regardless of the objective targeted, cannot, and should not be shouldered by senior female scientists alone, rather, it should be embraced by the scientific community as a whole.’”
In an open letter to the editor of Nature Communications, Leslie Vosshall, a neurogeneticist and HHMI investigator at the Rockefeller University, argues that the paper should be retracted. “The conclusions reached by the authors that being mentored by a female scientist is detrimental to career outcomes of young scientists, particularly female scientists, are based on flawed assumptions and flawed analysis. I find it deeply discouraging that this message—avoid a female mentor or your career will suffer—is being amplified by your journal,” she writes.
In response to these criticisms, the journal is reviewing the work.
“We believe that free inquiry and debate are engines of science, and welcome the review launched by the Editor in Chief of Nature Communications, which we think will lead to a thorough and rigorous discussion of the work and its complex implications,” the study’s authors write in their statement to The Scientist.
One thing both the authors of the paper and its critics can agree on is that the systemic biases against women in science will require systemic solutions. And one way that this bias can begin to be addressed, Sinatra says, is for committees involved in funding, hiring, and promotion decisions to take the citation index with a grain of salt. In short, because it is biased, “we cannot trust this measure. We need to make an active effort to not focus too much on this number or make choices based on optimizing this number,” she says.
Correction (November 24): Roberta Sinatra’s affiliation is IT University of Copenhagen, not the University of Copenhagen as originally stated. The Scientist regrets the error.