Public backlash against governance by algorithms has begun in Boston.
The recent uproar over proposed changes to Boston Public Schools start times should serve as a warning to policy makers and technologists who hope to use mathematics and computing to produce public policy that will impact the daily lives of human beings.
What’s the lesson here? In short: Computers, no matter how sophisticated, cannot solve political problems.
When the BPS announced it was offering $15,000 in prize money in search of an algorithm that would make its busing and school start times more efficient, it stated a clear goal: “freeing up funds to reinvest in schools and to improve the student experience.” Researchers at MIT’s Operations Research Group won the contest, designing an algorithm that shifts around 84 percent of school start times by moving many high school start times later and pushing many elementary school start times earlier in the morning. BPS seemed satisfied with the results, saying the new plan could save as much as $5 million in busing costs.
Particularly outraged are parents of elementary school students who may have to radically shift their schedules if school starts as early as 7:15 a.m. After sustained outcry from parents, teachers, unions, and even city councilors, the city’s justification for the changes has become substantially murkier.
Boston Mayor Marty Walsh released a statement reading, in part, “BPS is adjusting school start times with the aim to create a better, more productive learning environment for all students.” BPS Superintendent Tommy Chang said, “Start times were changed due to an abundance of research that shows student outcomes improve when secondary school students start later and elementary school students start earlier.”
So which is it? Did the Boston Public Schools adopt an algorithm coded to look for ways to enhance the learning experience? Or was the goal primarily to save the district money, as the BPS website said when it initially advertised the contest?
As mathematician Cathy O’Neil has written, algorithms are simply the product of our priorities, sketched out in code. That means that at its heart, the debate over adjusting school start times is a debate about our priorities as a city.
Of course, those priorities are not about math at all, but rather about politics — decisions about which city departments get funded and which get cuts. It’s perfectly reasonable for 21st-century governments to seek out algorithmic assistance to maximize efficiency or to predict how a policy will impact residents. But algorithms should never be used to paper over political or institutional deficiencies.
The new, algorithmically-generated school start times have left Boston parents asking whether the city is sacrificing their families’ needs to save public dollars.
We can imagine different scenarios producing vastly different results. For example, the city could crack down on the rate of Boston Police overtime expenditures, which totaled nearly $60 million in 2016. If city leaders reduced police overtime spending by as little as 8 percent a year, they could send students to school at developmentally appropriate times for all ages. Such a plan would enable high schools to start later without forcing elementary schools to start at 7:15, putting many families in a difficult situation. Doing so would require making different political decisions, creating different budgetary winners — and losers for whom no algorithm is likely to soothe the pain.
Algorithms, well designed, can solve problems and make things better. The important question is: Better for whom? That question needs to be decided in a transparent political process; otherwise, the algorithm will efficiently solve the wrong problem.Kade Crockford is director of the Technology for Liberty Program at the ACLU of Massachusetts, and Joi Ito is director of the MIT Media Lab.