AAAS Hosts Demonstration of New Diversity-Supporting Software

A computer scientist visiting AAAS demonstrated his innovative software program designed to aid college admissions officers in maintaining diversity while avoiding quotas or point systems that have been struck down by the U.S. Supreme Court.

The program, called Applications Quest, is being tested in a pilot program by a half dozen institutions, according to software designer Juan E. Gilbert, an associate professor of computer science and software engineering at Auburn University. One leading business school is expected to adopt the program permanently, he said, and input from other users has been positive. Gilbert said a privacy agreement prevents him from identifying the schools involved in the pilot study. He discussed his work during a 16 May seminar sponsored by AAAS’s Center for Advancing Science & Engineering Capacity.

Applications Quest groups a school’s applicants into numerous clusters, each containing students with similar backgrounds and qualifications. The students are grouped using a broad tally of common characteristics rather than a single criterion such as race or family income. For example, Gilbert said, it is rare for a single cluster to contain predominantly African-Americans.

While the students within a particular cluster are competing with others of similar background and interests, each cluster is substantially different from the rest. By selecting the top candidates in each cluster, the school can build a diverse pool of students, Gilbert said.

While the computer program recommends which candidates to choose from each cluster, the final choices are up to admissions officers, Gilbert said. They can include more subjective criteria in making those choices, including student essays or service activities that can be rated by application readers.

In choosing the clusters, the computer gives equal weight to various aspects of a candidate’s background, including GPA, SAT scores, home state, gender, family income and race. Admissions officers can choose to give one variable more weight than another, Gilbert said, although adding more weight to “race” would defeat the purpose of the program.

Once the data are entered, the computer program can compare attributes of 15,000 applicants in a few hours. While Auburn University has applied for a patent on the algorithm that does the work, Gilbert said Applications Quest is being offered to schools for free as an open-source program.

“It is a tool,” Gilbert stressed. “It does not replace people. It is a tool to assist them in their job.”

Gilbert, who is African-American, said he is passionate about giving others the same educational opportunities he enjoyed.

The challenge facing college admissions officers at selective schools is clear enough. There are many more qualified applicants than slots available. In a 2003 Supreme Court case, a white law school applicant at the University of Michigan argued that the school had rejected her in favor of a lesser qualified minority applicant despite her 3.8 GPA and a high score on the LSAT exam. The court held that schools may consider diversity in their entrance policies as long as race is but one of many factors considered. While the Court specified that diversity could be considered in college admissions, it did not specify how it should be used.

Michigan opted for a system to “holistically” evaluate admissions applications by ensuring that each one is read carefully at least twice. It requires a large staff of admissions officers and application readers at a substantial annual cost, according to Gilbert. Many schools do not have the resources for such time-intensive efforts, he said. The Applications Quest tool offers an alternative.

“This is a very exciting tool,” said Daryl Chubin, director of the AAAS Center for Advancing Science & Engineering Capacity. “It goes well beyond undergraduate admissions.”

He said Gilbert’s software program could be used to screen applications for professional schools, graduate schools as well as job placement for faculty or other professions. Gilbert said it likely could be adapted as well to screen applications for graduate research fellowships.

One of the participants in Gilbert’s pilot program is a graduate psychology program at a large public university. There were 270 applicants for 17 slots. Psychology department officials said that Asians were under-represented among their students. Gilbert’s program produced clusters that resulted in three Asians being recommended for admission, the same number as the department eventually offered to admit.

Gilbert said that his tool gives admissions officers a quantitative method for simultaneously comparing numerous attributes on thousands of student applications while also allowing them the final say in t complains about being passed over, he said, the officials can show that the student was in direct competition with others of similar abilities and background within a specific cluster of applicants.

Ernest McDuffie, deputy director of an Office of Naval Research program aimed at bringing thousands of new science and technology professionals into the Navy workforce, said he was intrigued by Gilbert’s presentation at the AAAS. He said use of clustering in the analysis of large sets of information makes sense. He noted that ONR is using similar computer methods to find common threads in diverse research reports that might not otherwise be apparent.

Earl Lane

19 May 2006