19

In the upcoming Fall (Fall 2019), I will be applying to CS PhD programs at some of the top research universities in the United States specialising in artificial intelligence.

While I cannot speak for other fields, I know that in my field (artificial intelligence), there has been a major push to increase the diversity of practitioners in the field. Here are a few articles about this diversity initiative:

  1. Article about lack of diversity in the field of AI: https://www.technologyreview.com/s/610192/were-in-a-diversity-crisis-black-in-ais-founder-on-whats-poisoning-the-algorithms-in-our
  2. Tweet thread by CS admissions chair at Cornell discussing diversity initiatives: https://twitter.com/davidbindel/status/992165303747461122?lang=en
  3. Tweet thread by PhD candidate at University of Toronto (a top school in AI) clarifying the importance of diversity in the field: https://twitter.com/leeclemnet/status/1040030107887435776

There are many other initiatives such as conferences at NeurIPS (typically regarded as one of the top 3 conferences in AI) that has workshops exclusively for members of underrepresented groups. I'm not sure what exactly the workshops entail, but here are the links to the relevant organizations: Black in AI, Women in Machine Learning, LatinX in AI.

The impression I get is that a lot of the talk around diversity tends to revolve around some key physical identifiers such as ethnicity and sex/gender; although perhaps I have not looked broadly enough at the full scope of diversity initiatives in the field and there may be initiatives that revolve around things like socioeconomic background, geographic background and national origin.

I can understand that when individuals of a particular ethnicity (such as African-Americans) and individuals of particular sex (such as women) are particularly underrepresented in the higher echelons of the field, people begin asking questions as to why that is; since, after all, ethnicity and sex should not be predictors of success in the field yet they clearly are.

Question

I'm an African student studying at a university that is ranked in the top-10 globally on both the Times Higher Education and QS rankings, and am planning on submitting some PhD (CS) applications to some of the top unis in the US in the upcoming fall. Given the preoccupation with diversity by practitioners in my field (in both industry and academia), I'd like to know to what extent my ethnicity will factor into the admissions decisions of committees in the US.

  • 8
    So is the question "are you concerned you will be the token gesture" or "will you be treated unfairly ie provided with less during the programme" ? – Solar Mike May 15 at 12:11
  • 7
    I suppose mainly the former: "will I be the token admit that is used to improve diversity statistics?" – UnchartedWaters May 15 at 12:23
  • 19
    @UnchartedWaters - this is precisely the problem (i.e., one major problem) with "affirmative action" as it is practiced in the US: it marks all persons who meet one of the affirmative categories as a recipient of the "action" regardless of that person's own wishes about the matter. (And seeking "diversity" is just the more modern terminology for "affirmative action" and applied to more selected groups of people. One might even think the name change happened in order to cloak the stigma associated with "affirmative action".) – davidbak May 15 at 16:09
  • 6
    It's important to note that a lot of action around these sort of things have to do with explicitly countering biases, rather than admitting Black people for the sake of admitting Black people. Research has shown that people with less typically Anglo names are passed over sooner in hiring, and that men and women students often underestimate their women colleagues. Part of the goal is that women and minorities get the fair evaluation they deserve, rather than being admitted to fulfill quotas. – Azor Ahai May 15 at 16:56
  • 8
    @davidbak your last sentence describes the problem with euphemisms, which leads to the Euphemism Treadmill: Term[n] is considered bad, so Term[n+1] is coined to replace it and remove the negative connotations. But soon enough, people realize it means the same thing, so Term[n+1] is considered bad, and Term[n+2] coined to replace it... Lather, rinse, repeat. Each cycle takes less time, because people have become trained to recognize what's going on. Example: "crippled" => "handicapped" => "disabled" => "differently-abled". – Monty Harder May 15 at 18:27
35

If there is one thing that academia is particularly bad at, it is transparency in decision making. Why do some people get admitted and not others? Why are some people hired and not others? Why are some people promoted and not others? From the bottom to the top of academic hierarchy, you'll be hard-pressed to ever get a straight answer, and answers you do get are always suspect because the nature of human decision-making is itself complex and opaque. There are plenty of reasons for this, from self-protection from criticism and gaming of the system, to rapidly changing criteria based on past experience and argument, to the fact that so much of the decision making is based on subjective ratings provided by individuals (often down to individual faculty members deciding if they want to work with a student).

To cut to the heart of your question, while it is not any sort of prescribed standard at the graduate level, you can safely assume race, ethnicity, national origin, and gender will be in some way be "taken into account" by some committees and individual professors. In many states this is actually illegal, explicitly forbidden by law - you can still assume it will sometimes happen anyway. On the other hand, some people will simply ignore demographic factors to the extent they are physically able, regardless of what they are told to do. As Universities have a history of avoiding transparency in admissions, you are extremely unlikely to ever know when it was considered, how much effect it had, whether it was positive or negative, and whether it ended up materially effecting the decision to admit you or not. You cannot know, they will basically never tell you, they will tend to go out of their way to avoid writing it down, and it is common for many people in the process to not even know with any certainty what factors ultimately "mattered".

However, from what I've seen on the inside and the outside of academia and from talking to fellow academics, you can safely assume at the graduate level that other factors matter way, way more. It is simply too high-stakes a decision for most schools to be altruistic, and they tend to steadfastly refuse to intentionally admit people who are utterly unqualified simply because of some demographic fact about them. The vast majority of Universities will only admit you if they think you will do well there, because having a bunch of under-performing students who are radically unqualified to pursue research and teaching is an albatross around the neck that few graduate schools will willingly tolerate. This is even more true at the PhD level, because they aren't cashing in on tuition (which is heavily marked up for international students, because money), so performance is king.

From talking to people working on diversity initiatives, I'll also give you another important insight: the diversity initiatives you hear about are not actually a preoccupation of the majority of faculty and staff, for good or ill. According to people working on actually making these initiatives successful, they report that the vast majority of people mostly want to ignore the issue and focus on other things, and so they work so very hard to make the initiatives public and heavily seen because otherwise they feel they will be unable to make any changes and they will be crushed by the majority weight of indifference mixed in with a heavy dose of active hostility to their efforts. Higher ed can look monolithic from the outside - but on the inside it is highly factious, and what appears as consensus is often a thin facade.

Because of all the above, the only advice I think makes any sense is treat the process as any other student should: apply broadly, see what offers you get, and talk with prospective advisors about what they are interested in. You are welcome to ask them if they were involved in your admission decision and then ask them why they wanted to admit you; I did that, and while it can feel a little awkward at first, a little bit of humor makes it easier and most professors when questioned directly, sincerely, and politely gave very direct answers. Some even said what parts of my profile they thought were a concern, and what other parts they thought would be a strength. Everyone gave wildly different answers, there was no consistency - no two people see the same person the same way.

Once you've got all your offers and finished talking with people, you can make your decision then on what seems like the best opportunity for you.

  • 3
    I agree with this answer, and I upvoted it, but I'd add that, although graduate programs won't admit genuinely unqualified students, there is a real question about the influence of demographic factors when the number of qualified applicants exceeds the number that can be admitted. – Andreas Blass May 16 at 2:30
  • 1
    @AndreasBlass Certainly, and I agree - but it still seems to reduce to "as an applicant you cannot do anything about it and can't know what net effect it had". It seems to me something that is important as a decision-maker, policy-maker, researcher in the area, engaged citizen, etc; for an individual student trying to navigate our byzantine world, it seems immaterial. – BrianH May 16 at 12:47
  • +1 I think this is the right answer. I might quarrel with your assertion that academia is particularly bad at transparency. Corporate hiring may be even less so. – Ethan Bolker May 16 at 13:31
  • 1
    @EthanBolker This is true. However, corporate hiring isn't near as clique-ish as academia, and there the bottom line matters way more than the appearances. Being half-scandinavian and half-latino made my life hell when dealing with some people from "top-10" universities, to the point of having to hear from a professor that they couldn't afford time to talk to a "jaja". On the flip side, I never had similar issues on interviews. YMMV. – T. Sar May 16 at 16:49
8

In addition to underrepresented racial and gender groups, some other diversity goals that many universities often care about are geographic diversity (not everyone coming from NY and CA) and admitting more first-generation college students.

I want to first echo what everyone else said, which is that schools don't admit people who they don't think will succeed, and if you're admitted they think you're a good fit. Second, the idea that if you put race and gender considerations aside then the default would be some kind of "pure meritocracy" is a fiction. There's a huge amount of randomness in students life experience and opportunities, in how good your letter writers are in writing letters, in who happened to be in the admissions committee that year and which letter writers they know. Schools also might want to balance between fields, and so if AI happens to get too many applicants at some school some year it might be harder for you than for someone in a different subfield. (And at the undergraduate level, there's a huge advantage given to legacies and athletes in obscure sports, both of which skew very rich and white.) At the graduate level probably the biggest of all, is there's a huge advantage given at state schools to American citizens and permanent residents (whose tuition waivers are significantly cheaper for the department to pay). If you're not a permanent resident, the disadvantage that causes is going to wildly outweigh any advantages that diversity considerations could possibly give.

Finally, I want to say that there are lots of simple practical reasons that diversity in hiring and admissions is practically important for departments. One of the most important ways that we are evaluated by the school is in attracting and retaining majors in our degree. On the whole, students are less likely to pick a major if they don't see any faculty or TFs with similar backgrounds to them. The African-American faculty I know are inundated with students who want to meet with them so that they can talk to a faculty person who understands their experience. Departments who don't have enough women faculty or don't have any African-American faculty are unable to do as good of a job of meeting the practical needs of their students.

5

I'll answer as far as how I perceived this working in the PhD program I graduated from, which was not in CS/AI.

A) No one was admitted who was unlikely to succeed in the program. No one was admitted who didn't belong here on merit in order to fill some sort of quota. No token admissions.

B) In every applying class, there were more applicants that the program wanted to admit than the program had funding to accommodate. Some of those students could be supported directly by a lab and join even if the program was out of money, but they would not have a year to rotate and test out different labs for a good fit.

C) Some additional funding was available to students from under-represented groups, however, only available to US citizens/permanent residents. Funding sources for international students were more limited, and they typically came either with funding from their home country or joined a lab directly that funded them.


In summary, the main way that ethnicity could matter was by funding for a student to join the program as a fellow/trainee rather than directly funded by a lab, which would give an opportunity to rotate in labs for a year that they might not have had otherwise. This was only even relevant for students who were already desired by the program on merit. As an international student this wouldn't have applied to you. Other programs may have different approaches that apply to international as well as domestic students.

  • "No one was admitted who was unlikely to succeed in the program." - this alone doesn't mean no discrimination. All who are admitted are qualified, but in a hypothetical case where there are 10 places, with 8 blacks and 100 whites qualifying, and 8 blacks and 2 whites were taken in, you would fulfill the "no unqualified for the sake of a quota" requirement, but would still be unfair (unless all 8 blacks were more qualified than all the others) and would potentially reject some candidates who would otherwise be even more qualified. – vsz May 17 at 6:10
  • @vsz See 6005's answer; your scenario is quite unlikely (simply, it does not happen). But frankly at that point, for PhD admissions, there is no such thing as more or less qualified except some particularly exceptional cases. – Bryan Krause May 17 at 12:24
  • I wasn't talking about how likely it is to happen, that's why I exaggerated the numbers in the example a bit. My point was that "No one was admitted who was unlikely to succeed in the program." does not, by itself, guaranty any fairness. – vsz May 17 at 12:31
  • @vsz But it does account for OPs concern about being a token admission. That does not happen. – Bryan Krause May 17 at 12:46
  • It does happen, there are clearly documented cases, and there are successful lawsuits which prove it. Maybe not at your department in your institution. (Please note that I'm in no way inferring that it was the case in the OP's situation, I was only countering the answer's argument of such things never ever happening) – vsz May 20 at 6:07
5

I agree with BrianH's answer, but wanted to add one thing. Everyone else seems to be assuming that if there is any bias in the admissions process, it would be in favor of underrepresented applicants.

This is the opposite of the conventional wisdom. In fact, it's well-documented in certain settings, e.g. applying to a job in the US, that when people make decisions about who is "qualified" or not, they are likely to be biased against women and underrepresented minorities. (This bias is often shown just based on the applicant's name, or other biographic details -- not a picture.)

I don't know to what extent this subconscious bias applies to academics at top research universities, but from personal experience I am guessing it really does, for some professors. You can safely assume:

  1. Some professors reading your application will show subconscious bias against your application based on your name, country of origin, or other demographic details (disfavoring women and URRM applicants)

  2. Some other professors reading your application may consciously favor your application because they wish to increase diversity in the department, and admit more URRM or women applicants.

  3. Either way, it is unlikely to be the biggest factor in the decision (see BrianH's answer).

  • 1
    I agree that this is an important point to make, but note the comment by the OP: "I suppose mainly the former: "will I be the token admit that is used to improve diversity statistics?"" i.e. that the OP was most concerned with token admission (which should probably have been edited into the original OP) and that's likely why the answers have an under-emphasis on biases against underrepresented applicants. Instead, they are focused on reassuring OP that they will not be admitted merely because they fill a quota. – Bryan Krause May 16 at 17:23
  • 1
    @BryanKrause But if it's true that there is widespread implicit bias against underrepresented minorities, then the OP's worry is unfounded. Due to the bias, giving some positive weight to ethnicity is merely counteracting the implicit bias, not unfairly admitting someone who is less qualified. The other answers are misleading because they suggest that giving positive weight to ethnicity is a form of unfair assistance, rather than a counteractive measure. – 6005 May 16 at 21:36
  • 2
    "then the OP's worry is unfounded" - Not necessarily. Both could be true: you could both have bias against minorities which leads to rejection of qualified minority candidates, and have corrective measures put in place to increase diversity that actually just select random (potentially unqualified) members of that minority to fill a quota. OP said they were most concerned with being admitted just to fill a quota, so that's what the other answers addressed. – Bryan Krause May 16 at 21:43
  • 2
    @BryanKrause Good point. I should have said, "then the OP's worry may be unfounded" (it depends on if the corrective measures are just to weight race a bit, or selecting candidates at random to fill a quota). But I think implicit bias is relevant and it may change the way the OP sees the situation. This is also very standard justification for affirmative action, so I was surprised that no one pointed it out. – 6005 May 17 at 0:56
4

Most (not all) US universities are in favor of diversity in the student body. However, this view is highly controversial in some circles and is actively attacked in the courts.

However, the purpose of the US notion of Affirmative Action was to try to ameliorate the effects of poor schools provided to minority and poor students along with active discrimination faced by such students as well as that faced by women in many fields. As such the rules are intended to benefit those who have grown up in the US but have been disadvantaged in many ways along their path to higher education.

The dilemma is that we have a belief in this country that everyone should be treated equally and haven't really found universally accepted ways to account for the fact that it isn't a reality.

But if you are a citizen of an African country rather than an American of (partial) African descent, then the rules, such as they still exist, won't really apply to you. The effect at the doctoral level is small in any case. The desire for a diverse student body is balanced by the desire to choose the "best" (most prepared) candidates. So, don't expect any strong bias in your favor. Hopefully you won't find any against you.

My advice is to assume that there will be no effect at all and to stress your qualification in any applications. Why can you be expected to be a success in your future studies? What is it in your knowledge and work ethic that makes you an excellent candidate? Every student needs to make that case of course. I think you are safest to assume it is the same for yourself.

I'll note that diversity of students based on their country of citizenship is also valuable to a university, as it often represents a diversity of viewpoint based on history and different educational systems. Universities do tend to value that when all else is equal (even if approximately). This can be especially valuable when questions of ethics (and similar) arise, even in a STEM field.

So, a big part of the desire for diversity in US colleges and universities is a desire for diversity of viewpoint, not just an ethnic (or gender) diversity. It might be hard to untangle the two in practice, of course.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.