Friday, November 15

Letter to the editor: Public perception of predictive policing is wrong, it can help reduce crime



Correction: The original version of this article incorrectly stated that Andrea Bertozzi is the Director of Applied Mathematics department. In fact, Bertozzi is the Director of Applied Mathematics program.

This post was updated May 21 at 11:28 a.m.

Dear Editor,

Earlier this month, Opinion columnist Lena Nguyen criticized the character and reputation of my colleague, Jeffrey Brantingham, an anthropology professor, in an Opinion column titled “Predictive policing algorithm perpetuates racial profiling by LAPD.” The piece maligned the nature of his work and contributions to the public – joining other attacks on his work in the past few weeks.

There is no merit to the accusations against Brantingham. There is a lot of misinformation and factual errors on this matter.

I am a faculty member in the mathematics department at UCLA and a member of the National Academy of Sciences. I have been the director of the applied mathematics program since 2005. For more than 15 years, I have been collaborating with Brantingham on the development of quantitative models and methods for social science applications, including, but not limited to, crime modeling. I am an investor in the company PredPol and have one publication with Brantingham related to the implementation of its software.

Information about our work is published in more than 15 peer-reviewed publications, written with roughly 25 doctoral students and postdoctoral members in mathematics, more than 10 undergraduate researchers, and faculty in statistics, computer science and criminology at UCLA, USC and UC Irvine.

The use of data algorithms for making decisions in our everyday lives has grown exponentially. Much of this occurs inside high-tech companies with proprietary software. Our team at UCLA believes in putting ideas into the public domain and in transparency of our work. We believe in peer review in leading academic journals. A rigorous study of the effectiveness of the PredPol software, published in the Journal of the American Statistical Association in 2015, concluded it led to significantly lower crime rates over a 21-month period.

I encourage all interested parties, especially those who are signing letters denouncing Brantingham, to read the relevant scientific literature. If you do not understand the technical writings, come and ask us and learn more about the methodology. I have read articles about predictive policing that have significant misinformation. I am concerned these pieces are driving the recent discourse.

I welcome a constructive discussion about the best practices for crime data analysis and data science in general.

Andrea Bertozzi

Distinguished Professor of Mathematics and Mechanical and Aerospace Engineering

Director of Applied Mathematics program UCLA

Share on FacebookTweet about this on TwitterEmail this to someoneShare on Google+Share on Reddit

Comments are supposed to create a forum for thoughtful, respectful community discussion. Please be nice. View our full comments policy here.

  • ANDREA BERTOZZI

    The title of the letter “Public perception of predictive policing is wrong, it can help reduce crime” does not represent the content and misrepresents the point of the letter. I did not say or agree to the choice of this title. Another clarification is that I work in the Mathematics Department. There is no Applied Mathematics Department on campus.
    Andrea Bertozzi