This story is over 5 years old.


Behind the Color-Blind Diversity Algorithm for College Admissions

Computer scientist Juan Gilbert explains how it's possible to have diversity without preference.
Image: pressmaster/Shutterstock

In April, the Supreme Court upheld Michigan’s ban against public universities in the state using race-based affirmative action in admissions. It’s the latest in a series of decisions whereby states have banned the use of race-based policies or schools have voluntarily abandoned them since former Florida governor Jeb Bush launched the One Florida initiative in 1999. States that have forbidden racial preference have seen a drop in black and Hispanic students.

There’s a movement to find race-neutral alternatives and try to keep up diversity by adopting class- and geography-based affirmative action, given the intersections of race, class, and place. One man, however, has proposed an algorithmic alternative. Juan Gilbert, a computer scientist and researcher at the University of Florida, has invented a software tool called Applications Quest that encourages diversity without giving preference to race. Two schools, Clemson and Auburn University, already use it in their admissions. I spoke to Gilbert about how it works.


MOTHERBOARD: What motivated you to create this software?
JUAN GILBERT: I created it in response to the 2003 Supreme Court decisions at Michigan on affirmative action. The motivation was the fact that at the end June 2003 both of the sides claimed victory, and at that point I knew that no one had won and this problem hadn’t been solved, which is evident by the fact that we are still dealing with it today.

How exactly does Applications Quest work?
If you look at what all the legal decisions have been, they say you cannot give preference—it says no preferential treatment. None of them say you cannot use or consider race, gender, or ethnicity. They just say you can’t give preference, but people misinterpret that as you can’t use it and the reason being is because no one knows how to consider it without giving it preference.

When you have more qualified applicants than you have available slots, you are going to turn away people who are qualified. 

On top of that with the recent case in Texas where the court is saying the onus or the burden of proof is on the university—that they considered other alternatives that are race-neutral and they have not been able to increase racial diversity, and they have to be able to show that they did not give preferential treatment. That’s impossible from their perspective to do, but that’s exactly what the software is able to do.

The software actually uses race, ethnicity, gender, [and] whatever attributes you want from your application in such a way that I can prove that none of these attributions are given preferential treatment; that they are treated equally, so what that means is a truly holistic approach. Holistic means you are making a decision based on multiple attributes. No single one of them is determining, and no one knows how to do that but this software does exactly that.


Does it use an algorithm?
That’s exactly right. It uses an algorithm that I wrote that determines the difference between two applications. Imagine if I gave you two applications and they were identical. A human being can say those are zero percent different and 100 percent similar, but what happens if you change the GPA on one and the major on the other? What’s the difference between a physics major and an economics major, or an African American and a Hispanic, or any other two attributes? What my tool does is accurately measure the difference between any two applications on a zero to 100 scale.

So what does that have to do with diversity? Well, what it does is, if you give me a group of qualified applications, I can actually compare each application with every other application, and once you do that, it can put them in groups or clusters based on holistic similarities. Those groups or clusters represent statistically diverse applicants and then the software makes a suggestion based on each cluster of one application to admit. It works and it’s compliant with all the legal decisions out there.

How does this fit into the debate going on right now about if race-based policies are the best way to give equal opportunity? With your software tool, are things like class, geography, and education levels of the parents taken into account?
They can be. I do not make the determinations of what attributes are used. The software is designed so that any admissions officer or whoever can determine which of the attributes are going to be used. They can capture legacy or first-generation college [attendees] if they have that information and can plug it in, and it will assess diversity based on those attributes.


If they get to point of saying you can’t consider race at all in applications, they would find something else to blame.

Why do you think there’s been a backlash or abandonment of these racially preferential policies? 
The whole debate we are having, it’s based on what I would call a capacity issue. To be honest with you, I don’t think it’s about race or ethnicity. The real problem is when you have more qualified applicants than you have available slots, you are going to turn away people who are qualified and when you do, they are motivated to sue you. If we didn’t have a capacity issue, we wouldn’t have affirmative action because everyone who was qualified would get in.

I think there a lot of people who aren’t racist but they feel like, my kid works hard, they’ve wanted to go to this university all their life and they didn’t get in. I don’t know why. So they start looking for a scapegoat. Race is an easy target for blame. If they get to the point of saying you can’t consider race at all in applications, they would find something else to blame—preference based on legacy or being an athlete. You have some racists out there but you have a lot of people who look around and think, "These people who don’t look like me must have been given a leg up." The woman who sued in Texas, though, her credentials weren’t that stellar. There were minority candidates that were better than her.

In the studies that I have done with Applications Quest, the myth that you have to sacrifice quality for diversity is just not true.

What studies have you done?
We go to institutions and they give me last year’s class of applications, and I can compare the group that they admitted by committee to the group that Applications Quest admits. We have over 30 cases, and in every single case, Applications Quest comes up with a more wholly diverse student pool.

We always have the same academic achievement level as the group the committee chose, and we do it in a fraction of the time. We can get greater diversity without sacrificing academic achievement, and we have proven that time and time again.