There’s been a lot of bloggage recently about a new study in the Proceedings of the National Academy of Sciences indicating bias toward male students on the part of faculty who thought they were evaluating an application for a laboratory manager. Half of the faculty in the study were given an application with “Jennifer” at the top, the other half one with “John” as the first name, and both male and female faculty rated the male student more highly, and would offer the male student a higher salary. Sean Carroll and Ilana Yurkiewicz talk about the study and the results in more detail.
So, without getting into the details of the study, let’s say I’m convinced by this that gender bias is a problem in hiring. Presumably, this would extend up to the faculty level, as well, where it’s even more important to take bias out of the process because the stakes are higher in a lot of respects. Now, one possible solution to this would be to try to make the hiring process blind– as has been shown effective in orchestra auditions, for example. So, if you wanted to make the faculty hiring process gender-blind (or race-blind, for that matter), how would you do that?
(This is not entirely a theoretical question, by the way, as there’s a non-trivial chance we will be hiring a visiting faculty member in the very near future, and a tenure-track job is not out of the question down the road a bit.)
Obviously, you couldn’t do a blind search all the way through, because at some point you’re going to invite a few candidates to campus for interviews. But would it be possible to do the search blind up to that point?
At first glance, it seems like a Hard Problem, because the materials that figure into a typical search are kind of difficult to anonymize to the necessary degree. For a faculty search, we ask for some statements from the candidate about their teaching and research, a CV, and three letters of recommendation. The statements are not necessarily gendered– though we occasionally get “As a woman in physics, I…” statements, those could probably be avoided. The CV is somewhat more problematic– while you could blank out the name at the top, the really important thing is the publication list, and there it’s a little hard to avoid identifying the author who is common to all the papers. It’s not a complete show-stopper– a lot of authors are identified only by initials, not full first names– but it’s trickier. Reference letters are the hardest of all, unless the people writing the letters are extremely scrupulous about avoiding gendered pronouns– in fact, we have sometimes had to resort to the reference letters for the purpose of identifying the gender of applicants with unusual names when we’re required to give some accounting of the number of applicants from underrepresented groups at the conclusion of a job search.
None of those are completely insurmountable– if the requirement were stated in the job ad, you could probably get the letter-writers to search-and-replace “Firstname” with “Dr. Lastname,” and the publication lists are primarily used to count the number of publications and the journals where they appeared, and only rarely does anyone look them up to be able to check the names. In the spherical frictionless world where support departments are supportive, you could probably have somebody in Human Resources transcribe any necessary documents and remove gender references. (We’ll pause here to give those who have had contact with real-world Human Resources offices a chance to stop laughing and collect themselves.) It’s not impossible, but the extra resources required make it seem like a bit of a hard sell.
I assume somebody must have thought seriously about this before, though I don’t have the time to Google for it. If there’s an effective and relatively simple way to go about this, though, I’d love to hear it. Even better would be a way to convince other faculty of the need for going through whatever additional hassle would be involved.
(The alternative, of course, is to be even more gender conscious, and make an affirmative effort to give extra consideration to applications from women (and other underrepresented groups, if we extend this to include other factors). I am somewhat skeptical that this would really be effective, though, as the judgments involved are inherently kind of subjective and prone to the semi-unconscious decision making that Kahneman talks about. And it’s not like we’re not already getting a lot of exhortations to hire from underrepresented groups– we’re aware of the general problem. I’d have more confidence in something that eliminated the possibility of bias at an earlier stage, shifting the effort involved to somebody else.)
(Obligatory disclaimer: Nothing in the above should be taken as a statement of institutional policy or a binding commitment to do anything in particular when next we hire. This is personal opinion and speculation, nothing more. If this post generates suggestions that seem workable and fit within existing institutional policies and applicable state and federal laws, I may try to implement them. My ability to force any kind of policy change is pretty much nonexistent, though, so I can not promise anything.)
Pulling gender pronouns out of rec letters would not necessarily solve your problem, because of gender bias in the way rec letters are written. See for example: http://www.ncbi.nlm.nih.gov/pubmed/19916666
I try to be aware of this when writing rec letters, and consciously think about the language and qualities I choose to bring forward in my letters.
It gets even harder if one of your applicants is a woman who, subsequent to her first publication, got married and changed her surname to her husband’s, as is traditional in some countries[1]. Even if she only uses initials in her CV, somebody is likely to notice that prior to a certain date F. M. Lastname was known as F. M. Maidenname.
For some non-Western applicants, the gender will not be obvious from the name. But these applicants will face discrimination in the form of a perception (occasionally justified, but frequently not) that they do not have sufficient command of English to teach effectively. (This is a separate issue from students who complain that they cannot understand somebody speaking fluent but accented English. I have much less sympathy for these students.) And not all non-Western candidates will benefit from gender ambiguity in the name; IME someone with an Indo-European native language whose first name ends with A is disproportionately likely to be female. Also, in some languages such as Russian, the surname will take on a different ending according to one’s gender, e.g., Dr. Ivanova will be female, but her brother would be Dr. Ivanov.
[1]For the record, as a Western male, I do not understand the persistence of this tradition. The woman gives up her name and in return she gets … um … what exactly? It’s one thing to do this in a place where it’s legally required (some countries have only ended this requirement during my lifetime), but other countries insist that the woman keep her name (e.g., China).
Yeah, I don’t think it’s possible to make it completely impossible to determine a candidate’s gender before the in-person interviews– a sufficiently determined member of a search committee can almost certainly work it out, given time. The question is whether there’s a way to present the necessary information that obscure the gender well enough that people won’t go through the effort.
For example, rather than listing papers with the full author list, you might be able to give the journal and a description of the contribution. Something like:
“Exciting new results from our experiment, ” First of five authors, published in Physical Review Letters (2011)
“Construction of an apparatus for exciting new results,” Second of six authors, published in Physical Review A (2011)
“Technical note about an aspect of an apparatus for exciting new results,” first of five authors, published in Reviews of Scientific Instruments (2010)
That’s probably enough information to determine the quality of the candidate’s research and their role in it (which is the relevant bit for a hiring decision) without immediately giving away anything that might bias the decision. Somebody who bothered to look up the actual articles would be able to figure out the candidate’s name and possibly gender, but it’d be an awful lot of work to do that with the 200-ish applications you get for a tenure-track position these days, and I suspect most search committees wouldn’t bother.
Are you hiring people who will already be known to your faculty? At a high level tenure track search, top candidates are often people whose research faculty already know of, in which case no anonymization will help.
David, that’s a problem for the high level search, but Chad’s talking about a visiting faculty position that might become tenure track in the future. I don’t think that’s as big a concern.
Chad, is it common practice not to read at least one or two of the papers on a CV at any point in the hiring process? Or is that just prior to a face-to-face interview, when you still have a list of 200 candidates?
We’re not really at the level where most of the candidates and their research would be known to faculty ahead of time. We do occasionally get candidates who somebody already knows, but it’s not an issue that comes up all that often.
I’m a little conflicted on that, because the fact that one of the faculty here knew me and my research already was a big factor in getting me this job. It’s the sort of thing that can shade towards “old boys network” kind of effects, though, so it might not be a bad idea to try to eliminate that in the early stages of a search.
As for the reading of papers, I don’t think anybody reads them at the first cut. Once we have a list of 20-ish, then people are more likely to look up the actual articles. Most of the cuts are made more on the basis of what they say in their statements than published articles, though.
Keep in mind, when we’re doing a search, we’re drawing applicants from a huge range of fields. I’m not really in a position to evaluate hard-core high-energy theory papers, for example, so I don’t usually bother trying to read them. What’s more useful and interesting to me is whether the candidate can make their research area comprehensible in the statements they submit, and whether they have thought about how to involve undergraduate students in the work. I assume that if they’re publishing in reputable journals relevant to their particular subfield, people who are more qualified to judge the details of the work have already made some evaluation of it.
Chad, are you saying that instead of listing paper titles one should just give descriptions? i.e. Were you being literal with the “Description of apparatus” part, or did you just put that there as a placeholder for the real title? Because if we start anonymizing to the point where people are allowed to editorialize instead of listing the actual bibliographical information, we’ll see all sorts of bad things. (And since men are considered more likely to brag than women, having people replace “Detection of X” with “First detection of an exciting and ground-breaking new phenomenon”, you’ll probably get disparate impacts.)
Also, I know a school that had the misfortune of hiring a few people who lied on the CV. At some point, the full bibliographical info needs to be there, because you’d be amazed what people will lie about.
Still, for a first pass, in principle I don’t see why a script couldn’t be produced to automate most of this: Replace “he” or “she” or [Name] with whatever gender-neutral prescription you like. Replace an author list with the position on the author list.
However, a lot of schools (perhaps to good effect, perhaps to no effect) now ask for applicants to include some statement about how they are sensitive to or active in promoting diversity. Those statements are very likely to elicit responses like “As a woman in physics…” or even “As a male, i try to be aware of my privileged position…” We can debate whether these statements actually help improve diversity in recruitment (it all depends on how the hiring committee uses it, in practice) or merely provide some fig leaf of cover aagainst lawsuits and/or bad PR, but as long as schools are asking for them they are essentially asking for candidates to identify themselves.
I was using those as placeholders for actual titles.
That would give the most neutral description of the work, and give anybody who really felt they needed to see the actual papers enough to find the article, via Google (that’s usually how I find arxiv preprints of stuff in paywalled journals, after all). It would make finding the full author list just enough work, though, that I don’t think most people would bother.
Here’s a related question to ponder: What about reviewing papers? If names were removed, would you be able to guess the identities of the authors on the papers you’ve reviewed recently?
I think I could identify the groups that most papers were coming from, either because I know who is doing what in my sub-field, or because of things like “Using the same techniques that we described in Reference 3…” Even if they removed the “we” from that, if they keep citing certain papers for methods and background info, and it is obvious that what they’re doing is an extension of those papers, it isn’t hard to conclude that they are the authors of those papers.
On the other hand, while I could identify the group, I wouldn’t necessarily know if the work was done by a male or female student or postdoc.
Alex @1425: Your question is entirely theoretical in my case–the journals I have reviewed for always tell me who the authors are. But even if they didn’t, the culture of this subfield is such that if the lead author works in the US (and in many cases this extends to other countries, particularly Canada/western Europe/Japan/Australia) I would still know with a high degree of confidence who the lead author is. That’s because our field encourages students and postdocs to attend conferences, and there are a couple of major conferences that almost everybody in the US (and many from elsewhere) attend. So in many cases I will recall the presentation by the first author when I review his/her paper, and it will be obvious who (s)he is.
I also agree with posters above who don’t like the idea of quasi-anonymizing a CV by identifying the candidate as author K of N and not giving exact titles. Especially when you are casting a broad net (as seems to be true of the putative visitor position Chad describes), different authorship practices of different subfields present challenges. High-energy particle people tend to give alphabetical author lists, so the candidate who is typically 1742 of 3003 will be at a disadvantage compared to one who is frequently first or second of a single or low double digit author list. Also, depending on who’s on the search committee, somebody who is often N of N (for N > 2) will be either at an unfair advantage or an unfair disadvantage if somebody involved is from a field which puts the authority figure at the end of the list (as some fields do but mine does not).
I think most physicists are aware that different fields have different practices for author lists. We might not be up on the details, but we know the gist. And I don’t know that saying “K of N” gives any more or less advantage than listing N names and putting the candidate’s name in Kth position. Either way, I learn that the candidate is K of N, but in one case I also learn the name.
From a practical point of view, completely blind faculty hiring is difficult–though I suspect the outcomes would be indistinguishable. An alternative to eradicating bias and overcompensating for bias, might be to simply recognize our own biases, and rely on scientific – like arguing and reasoning to help us make good decisions. Just because we’re scientists doesn’t mean we always think scientifically. But we should know how to strive for it.
One thing that has helped us is a practice where at each stage of the winnowing process, if our pool is less diverse than the original application pool, we have to articulate the reasons for our decisions in writing. It serves as a check on “Oh, look, our top 10 candidates all just happen to be white males!” and makes us actively think about our choices. To prevent bias bingo – no, this is not a quota system, sometimes the original list stands, but sometimes the articulation made people realize their original rankings were not based on actual qualities of the candidates.
It’s fairly common (though by no means customary) for CVs in math to list publications like:
(with Co-author1 and Co-author2) Title, rest of bibliographic info.
So my name never actually explicitly occurs in my list of publications.
(Of course this only works when it is near-universal for authors to be listed alphabetically and for author order to be meaningless.)
The problem with all of this is that you’re proposing solutions for the Easy Problem and ignoring the Hard Problem – which is that eventually the leading candidates will need to have interviews, and their gender (as well as other identifying characteristics) will become obvious. Unfortunately, there are still a lot of people on hiring committees who, however well-meaning they are, will not look at a female candidate the same way they’ll look at a male candidate. They wont realize it, but when the male candidate talks they’ll think he sounds like a promising future colleague if a little wet behind the ears, and when the female candidate speaks the exact same words they’ll think she sounds naive and inexperienced and not ready to be leading a research group. And they’ll tell the female candidate – very politely, very encouragingly – that they went another way but to stay in touch and feel free to apply for any positions that open up in the future.
At some stage, a personal conversation is necessary. However, the best solution is to make the list of applicants public knowledge, have them evaluated by external experts with no conflict of interest who rank the applicants in order giving justification and make this public as well. This is already the case in some countries. Anyone who gets hired but was not the best qualified will be obvious, and the institution will suffer as a result.