Mathematicians Urge Ending Work With Police

Nationwide protests against anti-Black racism and police violence have sparked demands and reckonings in several corners of higher education and academe.

The latest to join the fray is the discipline of mathematics. Ten mathematicians from eight universities and one company wrote an open letter last week calling for mathematicians to cease collaborating with police departments. The letter also urges math scholars to publicly audit influential algorithms and to embed learning outcomes related to ethics in data science curriculums.

"It's a political belief, that collaborating with police in any capacity contributes to white supremacist violence and oppression," said Tarik Aougab, a math professor at Haverford College and one of the letter writers, speaking personally on his motivations. "Really any collaboration between mathematics, which is something that I love and that I find extremely beautiful, and the institution of policing shouldn't happen."

The letter has 1,500 verified signatures from mathematicians, Aougab said, although that number is not limited to academic mathematicians. The letter has been submitted to and accepted by the Notices of the American Mathematical Society.

The letter writers take particular aim at "predictive policing," which involves using data and mathematics to predict where crime will happen.

"Many of our colleagues can and do work with police departments to provide modeling and data work," the letter said, noting that in 2016 the Institute for Computational and Experimental Research in Mathematics at Brown University held a one-week workshop on predictive policing.

PredPol, a major predictive policing company, says that its technology is being used to protect one out of every 33 people in the United States.

Opposition to predictive policing is not new. Academics have raised concerns in recent years about the mathematical theory underlying the technology and its tendency to create self-reinforcing feedback loops ("It is predicting future policing, not future crime," said a 2017 paper on the subject).

Although the group has gotten some pushback for not having personally worked with predictive policing mathematics, Aougab said that for them, opposition to the technology isn't based on finding errors in equations or formulas.

"It's not about our mathematical knowledge," he said. "We're using our mathematical knowledge and connections to further a goal that we inherently believe in politically."

Aougab said he personally feels police as an institution should be abolished, an opinion that is moving more into mainstream conversation.

This is not the first time mathematicians have waded into political action. In 2013, Alexander Beilinson, a University of Chicago math professor, wrote an open letter asking the American Mathematical Society to cut ties with the National Security Agency. In 2015, Data 4 Black Lives, a coalition of activists and data scientists, was founded to use data to "create concrete and measurable change in the lives of Black people."

In addition to ceasing collaboration with police departments, the letter calls for mathematicians to participate in public audits of any algorithms that could significantly affect people's lives. Algorithms that have been created to hire employees, sentence convicts or use facial recognition to identify criminals have been widely criticized for bias, inaccuracy and a lack of transparency.

ORCAA, a company run by one of the letter writers, Cathy O'Neil, evaluates algorithms for these traits, as well as for consistency and legal compliance. Additionally, Aougab said, mathematicians can use their expertise to explain to the public how algorithms can "bake in" existing bias.

"When an algorithm spits out an answer or a calculation, there's a tendency to just believe it," he said. "But there are no magic formulas. Any algorithm that exists has to be trained on existing data that is collected. If that data collection process is permeated by systemic racism, then your algorithm is going to perpetuate those forces."

The letter additionally calls for math departments with data science courses to implement learning outcomes about the ethical and legal implications of different tools. Though ethics courses might be common in biology or engineering departments, Aougab said they are nearly unheard-of in mathematics.

"There is a prevailing culture in mathematics that it is not concerned with the comings and goings of everyday life. It's abstract. It's pure. And for that reason, one doesn't have to consider ethics at all as a part of our work."

Part of organizers' motivations were subverting exactly that belief.

"Mathematics is very political," Aougab said. "And it's being used actively right now for oppression."