Colleges Are Using Big Data To Predict Which Students Will Do Well -- Before They Accept Them

Wrote this draft weeks ago, but never posted! Here it is...

Here's the article.

Ok. I'm normally all for Big Data. Generally, if a hospital network sharing my health history with other healthcare providers leads to better care for the community and more personalized care for me, I'm all for it. If Google, Amazon, and Facebook want to watch my web history and read my e-mails to deliver personalized product deals to me, so be it - I like buying stuff.

But this article was one of the first times I actually felt uncomfortable with the implications for using Big Data. The basic idea is that data mining of current student performance metrics can better predict prospective student performance than human admissions teams or consultants.

This may be true. But the notion that performance by past and current students should be used to determine whether this student today is going to flourish at a particular university sounds dangerous. This fails to account for the changing landscape of elementary and high school education. If a high school works to better prepare its students, but those new applicants are plugged into an algorithm that compares their stats to those of students who didn't go through the same system, are we really comparing apples to apples? And what data can we actually crunch beyond test scores - certainly, admissions essays could be run through an algorithm, but can a machine both detect and react to humor, well-formed metaphors, extraordinary examples of risk-taking and hardship, more effectively than a human admissions officer?

More in line with the focus of this course, how does this technology trend impact the education marketplace? Big Data fundamentally changes admissions strategies, as well as the playing field for inter-university competition. Let's take for granted for a second that the technology can, in fact, identify the best candidates. Envision a scenario where the wealthiest institutions can afford access to the most advanced predictive technology and therefore get first dibs on these "top" students. Now the country's most promising raw talent is flooding not to the places that offer the best education for the money, but to the place that had the time and the money to figure out how most efficiently to collect the diamonds in the rough. This shifts incentives away from investments that would be most beneficial to the education system - like programs that deliver more affordable education opportunities to more students.

There are obviously arguments on both sides here - mining the data might uncover candidates that would not otherwise have ever been considered for a college education, which leads to a paradigm shift in the way we assess candidates and therefore competitive differentiation among schools. It certainly warrants discussion.