This week we’re looking at Paul Tough’s comprehensive book on the ways college leaves lower class Americans behind, “The Inequality Machine”. In a late chapter Tough checks in on David Laude, a chemistry professor turned administrator at University of Texas, and Laude’s efforts to bring up the school’s graduation rate:
Laude needed to figure out exactly which students needed help. It was one thing to identify the students who might struggle in a five-hundred-person chemistry class. Laude was now responsible for the entire freshman class –more than seven thousand students. He conscripted a team of data scientists in the university’s office of institutional research who specialized in predictive analytics. By deconstructing the records of tens of thousands of recent UT students, they were able to develop a statistical model that combined fourteen separate variables, from family income to SAT score to high school class rank to parents’ educational background into a single algorithm that could reliably predict an incoming student’s likelihood of on-time graduation. They called this new tool the Four-Year Graduation Rate Dashboard.
Other schools now use similar dashboards. Here is the dashboard in action at Georgia State University:
One of the first students to arrive in Buis’ office on a mild morning in March, just after spring break, was Nicolas, a thin, soft-spoken, sharply dressed freshman with a sweep of sandy blond hair. When Buis opened his file on her screen, a big rec circle popped up that read “Predicted risk level: High!”
My first reaction to this was, wow, predictive analytics are very much here to stay. What started as a tool of high-finance and baseball dorks has now made its way down to advisory offices at public colleges.
My second reaction was, hell yeah mathematics! Laude and Buis are the good guys, the people trying to help poor students feel as if their struggles are not determinative or an indication on their abilities. Their schools have accepted low-income students into the jaws of the middle class, and there is an immediate culture clash. They are using their analytics to identify young adults who might especially need the concern of supportive adults. Yes, this is the best case scenario.
My third reaction was, there sure have been a lot of people using predictive analytics in this book. The one that most caught my attention wasn’t from advisory offices, but admissions offices. What I have learned is that admissions offices hire data analysts to model the financial feasibility of admitting various students. How much tuition revenue would we generate if we admit so many students? How much aid would we be on the hook for? What would the classes average SAT/ACT score be? Will our US News rankings go up or down?
In the coming years, it will continue to get cheaper and cheaper to create predictive models. They’ll make their way into more and more industries. They are tools of efficiency, equality, inequality – in other words, they are tools of humanity. I think they’ll be used more for good than for evil. But who can say for sure?
Oliver Sacks documentary, free on PBS // Best of MF Doom // Flowchart of philosophical novels // Breaking Down David Bowie’s “Heroes” in the studio // Research: income and selective college attendance // 2021 best essay collections