A data-driven approach can retire the endless, unproductive battles over race- and class-based affirmative action.
In his column “‘Economic Affirmative Action’ Won’t Work” (Upward Mobility, April 23), Jason Riley critiques Richard Kahlenberg’s call for elite colleges to favor class indicators—such as parental income and neighborhood poverty—to restore the racial mix once achieved through race-based affirmative action. Mr. Riley worries that class preferences will dilute merit and mismatch underprepared students with unforgiving curricula.
Both men raise important arguments, but they miss the point: Choosing whom to admit—or hire or promote for that matter—isn’t a philosophical exercise. It is a classic statistical problem. Modern machine-learning tools can solve it far better than any committee armed with SAT scores, legacy flags and, worse, their gut instincts.
Evidence is clear that machine-learning algorithms can make better decisions than humans, especially when bias is present. A February 2017 study of judges’ bail decisions found that algorithmic recommendations could cut crime by up to 25% without increasing jail rates. Stanford University researchers showed that a computer-vision model matched or outperformed radiologists in reading chest X-rays. In my own work with dozens of firms, machine-learning-based “hiring co-pilots” have reduced regrettable attrition by 47%. If algorithms can keep communities safer, spot cancer earlier and put the right people in the right jobs, why not deploy them in college admissions?
Continue reading the entire piece here at the Wall Street Journal (paywall)
______________________
Roland G. Fryer, Jr., a John A. Paulson Fellow at the Manhattan Institute, is Professor of Economics at Harvard University, an entrepreneur, and co-founder of Equal Opportunity Ventures.
Photo by gordenkoff/Getty Images