Admissions by Software

Suppose colleges could buy software that would enable them to speed up the admissions process and admit classes that are more diverse—without fear of being sued by foes of affirmative action. Would institutions purchase? Can (and should) colleges outsource admissions decisions to software?

That’s what a professor at Auburn University is hoping to find out. Juan E. Gilbert, an associate professor of computer science, developed the software two years ago, and it attracted a brief flurry of attention. But what he didn’t have then were any clients to actually use the software. Now, Auburn—which tested the software last year—has announced that it will use it in actual admissions decisions next year. And while Gilbert says that Auburn is the first institution to publicly admit to using the software, he says that about five other colleges have used it, although in many cases not for an entire class of undergraduates. Gilbert also now has an outside investor, having sold a minority interest in the company to the owners of Cox, Matthews and Associates, which publishes Diverse Issues in Higher Education.

It’s unclear of course whether Gilbert’s software business will take off. But Applications Quest raises interesting issues about how colleges evaluate and measure diversity when they assemble their classes. Gilbert’s basic premise is that colleges want racial and ethnic diversity in their student bodies, that the Supreme Court’s rulings allow colleges to consider race and ethnicity only in certain instances, and that many colleges are afraid of being sued.

{snip}

The software works this way: A college decides how many applicants it will admit from that middle group. Then the college picks the criteria it will use to evaluate the applicants. These would generally include a range for acceptable SAT scores or high school grades, planned major in college, status such as first generation to go to college, intended major, race, gender and so forth. Then the software clusters all of these applicants into groups equal in total number to the number of projected admits. In doing so, the software groups applicants who share the same qualities together. Then the software selects one person from each group to be admitted, but this person is the applicant who is the least like others in his or her cluster. (Details are available on the company’s Web site.)

Gilbert said in an interview that this system offers key advantages. First, it removes bias from the equation, as an admissions officer can’t be accused of providing more or less emphasis to any criterion based on personal interests. Second, it assures that race and ethnicity are taken into account, but not given more weight than the Supreme Court would permit. “It does what we as human beings can’t do on a large scale,” he said.

{snip}

While Alderman said he is certainly following the debates over affirmative action nationally, that issue was not crucial to the interest at Auburn. The protection against lawsuits “is certainly nice to have,” but “for us, it was a more efficient way to achieve the same results.”

Roger Clegg, president of the Center for Equal Opportunity, a group that opposes affirmative action, said that while he thinks this software may be “less racially discriminatory” than the way some colleges have assigned points for applicants on the basis of race or ethnicity, concerns remain. Clegg said that the software is still racially discriminatory in that it considers race and ethnicity, and that the system was designed “in order to achieve particular racial results.”

Clegg said that the problems with the software are apparent if one imagines the same computer programs being used to keep black students out—as opposed to admitting them. Just as one would oppose the use of software to exclude members of minority groups, he said, the same should be true of software with a diversity goal.

{snip}