College Arms Race: WMD Version

Weapons of Math Destruction lifts the veil on the darker side of Big Data. In ten chapters of vignettes, the author Cathy O’Neil appropriates the acronym WMD bringing it very close to home to anyone touched by the new economy – a globalized technocentric world order. O’Neil has had a varied career. With a Ph.D. in Mathematics from Harvard, she started out as an academic at Barnard. But then the siren song of investment banking and hedge funds called out to her, and she joined D.E. Shaw as a quant. (Having a computational background, I was once tempted to apply to D.E. Shaw earlier in my academic career, but I think I’m much happier teaching college students.) Disillusioned by the financial crisis, she “was especially disappointed in the part that mathematics had played [and] was forced to confront the ugly truth: people had deliberately wielded formulas to impress rather than clarify.” Leaving the hedge fund, she has worked on illuminating WMDs, and also started the Lede Program in Data Journalism in Columbia.

The chapters include (1) a peek at predictive crime models that can create pernicious feedback loops, (2) personality “tests” on job applicants that lead to new automated forms of discrimination, (3) corporate efficiency schedulers that disenfranchise employees, (4) using credit reports as proxies for all manner of future predicted risk behavior, (5) large-scale social media manipulation, and more. If any of these topics interest you, I highly recommend reading her book in full. The two topics I will highlight in today’s post are Chapter 3 (“Arms Race”) and Chapter 4 (“Propaganda Machine”).

As a faculty member in a private liberal arts college in the U.S., I have grown increasingly concerned with the unsustainable arms race that has accelerated in earnest over the last 5-10 years. Holding down tuition is increasingly difficult as institutions fight for the small swath of students that will increase their prominence so as to keep themselves afloat. Administrative staff must be hired to support the ever-increasing facilities and amenities in an effort to be a full-service college. Increasing mental health concerns among students are putting strains on universities struggling to provide adequate support. The current strategy of discounting to draw more desirable students is financially unsustainable, and yet the arms race continues.

O’Neil begins this story in 1983 with U.S. News and World Report’s first set of college rankings. It says something that in 2017, the “rankings” have expanded to cover more types of institutions and programs, while its original news-magazine roots have died with the times. Top Ten lists have always been popular, but Big Data, with its whiff of being scientifically based, has turned these lists into a gargantuan beast of its own. The siren call of Big Data is strong, perhaps overwhelmingly so. With more data, we should be able to make better decisions. It is a strategy that runs well when making a pitch to administrators for more money; I’ve used it myself multiple times. The larger and more complex a system becomes, the more it needs to feed on Big Data to optimize efficiency. We humans might say to ourselves that we are the overlords of the system, but more and more are becoming enslaved. How exactly? Thanks to the correlation between increasing complexity and obtuseness, things are becoming less and less clear. That’s why O’Neil’s book is important in highlighting these WMDs.

Everything starts with a model. Ideally, and this is most true in the “hard” sciences, the feedback loop in developing a model is much more rigorous. You test the model with experiments, generate more data, and then refine the model using the new data. Rinse. Repeat. The models that work best have narrow and clearly defined outcomes. But many complex problems are “squishier” (O’Neil’s term), and the journalists tasked with ranking colleges were trying to measure something as nebulous as “education excellence”. O’Neil writes: “They had no direct way to quantify how a four-year process affected one single student, much less tens of millions of them.”

So what do most people do in this case? You pick a proxy. Now hopefully you choose a proxy that correlates well with what you’re trying to measure, and you provide appropriate feedback into your model. In the college rankings game the proxies that “seemed to correlate with success”: SAT scores, student-teacher ratios, acceptance rates, retention rates, alumni giving, and peer evaluations. With fame or infamy come problems. O’Neil explains: “As the ranking grew into a national standard, a vicious feedback loop materialized. The trouble was that the rankings were self-reinforcing. If a college fared badly in U.S. News, its reputation would suffer, and conditions would deteriorate. Top students would avoid it, as would top professors. Alumni would howl and cut back on conditions. The ranking would tumble further. The ranking, in short, was destiny.” In recent years we’ve heard a number of stories exposing institutions that attempted to game the system. What a sad state of affairs.

It gets worse. O’Neil explains that to establish initial credibility, the early lists needed to have the known “elites” on top. What made them special? High SAT scores, great graduation rates, excellent retention, strong giving from rich alumni, and name-recognition among peers. Sound familiar? Glaringly, cost of attendance was not included in the formula of the model. Perniciously, this led to gaming the system without needing to keep tuition down. In O’Neil’s words, “in fact, if they raised prices, they’d have more resources for addressing the areas where they were being measured.” Tuition has skyrocketed in the last twenty years – a correlation, at least, if not one of the causes.

As the arms race increases the gap between the haves and have-nots, the next WMD comes into play: the propaganda machine of for-profit higher education advertising. Big data has sharpened this model with deadliness in targeting the most needy with the taglines most likely to succeed in prying cash they don’t have from their wallets. Uncle Sam steps in with federal loans to make up the difference, up to a whopping ninety percent! O’Neil writes: “Anywhere you find the combination of great need and ignorance, you’ll likely see predatory ads… they zero in on the most desperate among us at enormous scale.” That last phrase is what defines a WMD, the enormous scale. Tuition is set to maximize borrowing at the limits of federal loans. But, that’s not so important if you’re selling the dream of upward mobility. Society’s veneration of the entrepreneur, coupled with well-chosen anecdotal stories, provide what may sound like the only hope of those in difficult circumstances. No risk, no payoff.

The details of this sad story are told in Lower Ed: The Troubling Rise of For-Profit Colleges in the New Economy by Tressie McMillan Cottom. Before she became a sociologist and professor, Cottom worked as a recruiter and enrollment specialist in the for-profit world. She eschews the simple tropes used to explain what is happening at the other end of the spectrum, far away from the noses of the “elites”. Her analysis is penetrating, and I strongly recommend her book if you want to know how and why Lower Ed began its thriving ascent. As they add a slew of graduate programs and post-baccalaureate certifications to their programs, their nimbleness is outpacing traditional public and private institutions of higher education. There is no simple way to assign blame, because this complex issue is part of a much larger ecosystem

Here’s a quote from the Epilogue that summarizes one of the main issues. “In the absence of social policy, public subsidies to Lower Ed become a negative social insurance program. A negative social insurance program is a market-based response to collective social conditions. Negative social insurance, unlike actual social insurance programs (e.g. Social Security), doesn’t actually make us more secure. It only makes our collective insecurity profitable.” Given the limited information I’ve seen in the current (U.S.) Administration’s higher education stance, the gap between the haves and have-nots will widen further. For-profits aren’t going to bridge this gap but instead will “perpetuate long-standing inequalities”. I’m not doing justice to Cottom’s careful argument; I recommend reading her book in full.

O’Neil closes her book by reminding us that Big Data is here to stay. “Predictive models are, increasingly, the tools we will be relying on to run our institutions, deploy our resources, and manage our lives. [But] these models are constructed not just from data but from the choices we make about which data to pay attention to – and which to leave out. Those choices are not just about logistics, profits, and efficiency. They are fundamentally moral. If we back away from them and treat mathematical models as a neutral and inevitable force, like the weather or the tides, we abdicate our responsibility. And the result, as we’ve seen, is WMDs that treat us like machine parts in the workplace, that blackball employees and feast on inequities… Math deserves much better than WMDs and democracy does too.”