CASMI Hosts Workshop on Data-Driven Policing

The Center for Advancing Safety of Machine Intelligence (CASMI) at Northwestern University hosted its first workshop entitled, “Best Practices in Data-Driven Policing” last month. The two-day workshop functioned as a forum for facilitating interdisciplinary conversations focused on the concerns and benefits of data-driven policing, as well as best practices for developing and implementing data-driven policing technologies.

“Any conversation with a wide range of people, like we had here, is the most productive. The way each one of these groups looks at the topics and issues is very different and that gives us a clearer picture of the potential consequences of using these technologies,” said Ryan Jenkins associate professor of philosophy at California Polytechnic State University and senior fellow at the Ethics + Emerging Sciences Group.

Presenters and participants were experts or practitioners in many fields and institutions including municipal law enforcement, litigation, philosophy, computer science, community advocacy groups, professional organizations, and technology development. The workshop reflected this interdisciplinary make-up with sessions focused on policing goals, feedback loops, civil liberties, and technology demonstration, among others.

Through presentations and subsequent discussions, multiple threads emerged regarding ongoing issues, practical solutions, and next steps. Jenkins spoke on one prominent theme, the difficulties in measuring success of data-driven policing technologies. “We just don’t have enough data. There is an abundance of data for systems to be built on, but we don’t have enough information on the systems themselves for evaluation or comparison purposes.”

Jenkins suggested one path forward would be the identification and publication of case studies on successfully implemented data-driven policing systems. Case studies could supply insight into roll-out methods, define minimum levels of efficacy, and be used to determine data gathering techniques.

The indispensable role of community was a persistent discussion theme that arose even before its dedicated sessions. Technologists, advocates, academics, and practitioners agreed that the ultimate stakeholders of data-driven policing systems are the communities in which they operate. The creation of community liaison positions within technology companies and government regulatory bodies was proposed and discussed at length. Liaisons could contribute to a multitude of solutions for issues including community endorsement, increasing transparency, and privacy and fourth amendment concerns.

The Best Practices in Data-Driven Policing Workshop represents the culmination of a three-year research project at Cal Poly, San Luis Obispo, and the University of Florida, “Artificial Intelligence and Predictive Policing: An Ethical Analysis.” The project’s co-principal investigators are Ryan Jenkins, and Duncan Purves, associate professor of philosophy at University of Florida.

One of the benefits CASMI sought in supporting the research project and hosting the workshop was the opportunity to refine its approach to assessing intelligent technologies. Data-driven policing is a complex and sensitive field and thus an opportune area to exercise CASMI’s “Evaluation Framework” for assessing the efficacy and safety of algorithmic systems relative to the goals of the domain in which they operate.

CASMI director and Bill and Cathy Osborn professor of Computer Science at Northwestern Engineering, Kristian Hammond, spoke on the significance of hosting this workshop for CASMI. “The workshop on Data-Driven Policing is a prime example of the kind of work that CASMI is designed to support. On one hand, predictive technologies utilizing machine learning have the potential to improve public safety by identifying the areas where policing is most needed. On the other hand, such predictions can lead to feedback loops and self-fulfilling prophecies. In this workshop, a set of diverse thinkers were able to walk through the processes and craft mechanisms that provide the benefits while avoiding the problems.”

CASMI sees these conversations as a starting-point for addressing these issues in data-driven policing and AI development as whole. The center aims to host more workshops and facilitate more collaborative research projects to bolster its mission of developing best practices for the evaluation, design, and development of machine intelligence that is safe, equitable, and beneficial.