Mathematical Sciences Research Institute

Home > About > News > MSRI in the Media > Show

Computer program tackles diversity in college admissions

  1. June 26, 2004
  2. Betsy Mason

BERKELEY - A mathematician, who wants others to have the same opportunities he had, has created a computer program that he hopes will be a low-cost answer to a problem presented by a Supreme Court ruling on college admissions.

Although generally viewed as an endorsement of affirmative action, last year's Supreme Court ruling that college admissions offices can consider diversity as long as race is not treated as a deciding factor left some fearing that minority enrollment might drop if schools don't have the staff, time or money to take the "holistic" approach the court ordered.

The University of Michigan spent $1.8 million implementing a new admissions system in order to comply with the court's decision. They hired five more full-time admissions counselors and 16 seasonal application readers so that each of the approximately 20,000 applications the school receives would get at least two thorough reviews.

While that price tag might not faze the University of Michigan, which has one of the largest endowments in the country, schools without such deep pockets might wonder if they can afford such detailed reviews.

"Had it not been for affirmative action, I don't think I would have had the education opportunities I did," said Juan Gilbert, an African-American mathematician who attended Auburn University in Alabama..

"Diversity could be lost because not everyone has the bank account that the University of Michigan does. So, I felt that I had to do something."

Gilbert, who presented his findings this week in Berkeley, tackled the problem by clustering algorithms -- step-by-step calculations that group similar items together. He developed a computer program that uses these algorithms to compare all aspects of an application -- such as GPA, hometown and race -- to every other application and gathers similar applications into groups.

The idea is that applications in any given cluster are more similar to each other than to the rest of the applicant pool. Because race is just one of the attributes considered, clusters are likely to include several races, just as they will have a range of SAT scores and a mix of genders.

"Diversity isn't necessarily race alone," Gilbert said. "It's everything -- it's the whole application. That's how I'm defining it."

Schools can then accept an equal number of applicants from each cluster. How these choices are made would be up to the individual school's admission officers, and this could be tricky.

Random selection would be the simplest approach, but schools are likely to want to use criteria such as essay answers or community service or other intangibles that can't be quantified. As long as race isn't the deciding factor at this point, however, the process should adhere to the Supreme Court's ruling and still create a diverse class, Gilbert said.

"If the University of Michigan had been using this (program) when they were sued, they could have sat the young lady (who filed suit) down, pulled up her cluster and explained why other members of that cluster were accepted ahead of her," he said.

Gilbert presented his ideas Wednesday at the 10th annual Conference for African-American Researchers in the Mathematical Sciences, held at Lawrence Berkeley Laboratory and co-sponsored by the Mathematical Sciences Research Institute in Berkeley.

"I think it's a very clever idea," said William Massey, a mathematician at Princeton University and one of the founding organizers of the conference.

"I've done admissions before, and it's so subjective," said mathematician Arlie Petters of Duke University. "It's great to have these tools to serve as quantitative means to guide you as you make human judgments."

In its basic form, Gilbert's program considers each bit of an application's information equally, but it allows the user to put more weight on any particular attribute or attributes. While this would give admissions offices more flexibility and control, it also leaves the door open for some abuse.

"I recommend giving everything equal weight," Gilbert said. "If you didn't want it considered, you wouldn't put it on your application."

Auburn University has applied for a patent for Gilbert's software, called Applications Quest, which is already generating a fair amount of interest in the business world. Employers could use the program to screen job applicants by putting a fake application for the ideal candidate into the mix, and then interviewing people who end up in the same cluster with the ideal.

But Gilbert's focus is maintaining diversity on college campuses. "I think diversity is more important than most people realize."