City Hall

Dallas Takes Up a New Tool to Map Crime Risk, but Reformers Worry It'll Reproduce Old Biases

Daniel Rockey
Dallas has begun a new era of data-driven crime targeting in neighborhoods.
In 2013, the Arlington Police Department began experimenting with a then-novel way of looking at crime. With money from a National Institute of Justice grant and in partnership with academic criminologists, Arlington used Risk Terrain Modeling, a data analysis tool, to combine existing crime reports with geographic information in hope of finding what environments contribute to property crime.

A second phase of the project was intended to create alternative ways of improving public safety, but personnel turnover in the department led to the city scuttling the project before the second round of funding came through. A report on the first phase’s findings in 2015, though, outlined seven geographic elements most associated with property crime: apartment complexes, schools, foreclosed properties, pawnshops, variety stores, convenience stores and gas stations.

That describes virtually every less affluent neighborhood in DFW, and the link between crime and income is well-established. But, the promise of Risk Terrain Modeling is that by tightly focusing on crime’s spatial characteristics and how they interact, it can help create maps that predict the risk of crime before it happens, and not just at a ZIP-code level, but block by city block. That could allow communities to intervene in other ways, through zoning and code enforcement, for example, and alter the environment rather than flooding neighborhoods with squad cars.

Nearly a decade later, as violent crime increases and residents demand police reform, Dallas has moved to make Risk Terrain Modeling a critical part of a larger, long-term strategic shift in how the city approaches policing and public safety.

That’s the hope, anyway, but some reformers are wary. They question who will oversee the program’s use, what data will be fed into it and how the city will ensure it doesn’t reproduce old biases. And they fear the modeling’s focus on place is an invitation to more gentrification in lower-income neighborhoods.

Developed by two Rutgers University academics to analyze crime data, Risk Terrain Modeling also has been used to examine the link between terrain and a broad range of phenomena — cholera outbreaks, traffic accidents, overdoses and pollution, for example. But crime remains the primary focus for U.S. cities picking up the tool.

Early this year, before this summer’s protests against police violence, The Dallas Morning News reported on the adoption of Risk Terrain Modeling by the Dallas Police Department. Three “risk maps” were featured in the article, outlining areas identified as high risk for violent crime and robbery. In February, Dallas Chief of Police Renne Hall told NBCDFW that the approach would be used to confront problems like “poor lighting and abandoned businesses,” but she conceded it may be used to deploy “targeted patrols” to identified areas.

click to enlarge
Changa Higgins is the leader of the Dallas Community Police Oversight Coalition, which campaigned to form the Office of Police Oversight.
Kathy Tran
Proponents of Risk Terrain Modeling appear eager to emphasize the environmental interventions and distance it from hot-spot policing, the practice of mapping where crimes have happened and using that to predict where best to station police. Dallas’ use of hot-spot policing in the past “led to an oversaturation of police and traffic stops —  the Dallas version of stop-and-frisk,” says Changa Higgins, leader of the Dallas Community Police Oversight Coalition, which campaigned to form the Office of Police Oversight.

 “We don’t want this to be used as a hot-spot policing machine” says Alan Cohen, head of the Child Poverty Action Lab, a nonprofit that conducts the technical risk terrain analysis for Dallas.

Former Assistant Chief of Police David Pughes was previously responsible for coordinating Risk Terrain Modeling efforts inside the Dallas Police Department. In late April, City Manager T.C. Broadnax announced the creation of a new department responsible for Risk Terrain Modeling, called the Office of Integrated Public Safety Solutions, with Pughes at its head.

“Some jurisdictions have sought to use Risk Terrain Modeling for heavy-handed predictive policing based on strict and basic logic of where crimes have previously occurred,” Pughes says. “But we can’t continue to utilize the police department in the same fashion. They are being asked to do too much and stretched beyond their expertise. Risk Terrain Modeling can be used to inform unconventional interventions that will ultimately reduce the number of police officers in high risk areas.”

As is often the case when dealing with issues of police reform, one of the challenges Pughes faces is overcoming mistrust.

“Efforts to lighten the burden of police officers should not fall under the same old structure,” City Council member Adam Bazaldua says. “I’m concerned about how data can be used in a manipulative manner or result in over-policing.” (He’s not being cynical. When Bazaldua introduced a resolution to decriminalize possession of small amounts of cannabis in Dallas in September, the police department produced a particular month of data that showed a significant number of marijuana arrests being connected to violent crime, but the correlation isn’t widely supported.)

The creation of the Office of Integrated Public Safety Solutions appears to be a concession to soothe doubts about the modeling. “This is an opportunity for a new paradigm,” Pughes says. “We need to differentiate between law enforcement and public safety.”

A common refrain among proponents is that Risk Terrain Modeling aims to shift the focus toward place and away from people. “The people come and go, but the problem locations of crime remain relatively the same,” Pughes says.

“But it doesn’t replace the human element. It actually requires local expertise,” Cohen says.

Community meetings are used to gather information. “The answers are clearer and simpler for those who know the area,” Cohen says. Cohen believes the approach can inform a number of unconventional interventions that improve quality of life.

An example is the renovation of the Five Points intersection in Vickery Meadow by the Better Block Project. In November 2019, the Morning News reported the project added “a plaza, potted trees, a pop-up container store, stage, swings and colorful walkways to beautify the neighborhood.”

In an email to the Observer, Cohen wrote that  “risk has lowered in the city blocks within where Better Block's environmental design interventions were enacted last year.”

How much of the change is attributable to the facelift is unclear. Better Block is but one effort that has emerged from Project Safe Neighborhoods, a collaboration with federal law enforcement that cited over 100 arrests in the broader area in 2018.

Ironically, the creation of the new Office of Integrated Public Safety Solutions puts Risk Terrain Modeling outside of the purview of the Office of Community Police Oversight, which was created in 2019 in response to the murder of resident Botham Jean by an off-duty police officer.

Although the new Office of Integrated Public Safety Solutions is not a law enforcement organization, Pughes and the office are also responsible for the expansion of the Right Care program, which pairs social workers with police patrols, and the Violence Interrupters initiative, which deploys trained civilian de-escalators instead of police officers to defuse potentially violent situations.

“I’m concerned that the new office could be siloed away from the Office of Police Oversight. There needs to be some sort of oversight of the [Risk Terrain Modeling] process” Higgins says. Higgins has been working with the Child Poverty Action Lab for several months to design and conduct community outreach sessions to inform the broader strategy for Risk Terrain Modeling.

While Pughes has committed to working with the Office of Community Police Oversight, he confirmed there is no formal oversight commission for the work conducted by his new office as it relates to law enforcement or for the implementation of Risk Terrain Modeling.

This fact, as well as the broad range of applications of Risk Terrain Modeling, raises other red flags for people like Dr. Katy Mead, who works with local activist group People Against The System, which conducts weekly educational sessions on issues related to systemic oppression.

On Aug. 21, Mead was one of a dozen protesters who gathered outside Broadnax’s home as a part of the call to reallocate significant amounts of funding away from the police department. After an initial and somewhat confrontational exchange, Broadnax invited the group into his backyard for a conversation.

click to enlarge
Activists Ashunti Radford (left) and Katy Mead question whether the new system of assigning risk to neighborhoods mirrors redlining, the historic and systematic denial of investment and services to Black communities by federal government agencies, local governments and housing lenders.
Kathy Tran
Mead questioned whether the “risk maps” would simply highlight areas in Dallas that had faced redlining, the historic and systematic denial of investment and services to Black communities by federal government agencies, local governments and housing lenders.

In 2019, researchers at St. Louis University used Risk Terrain Modeling and other geospatial assessment techniques to show that the risk for robbery was associated with areas in St. Louis that were subjected to racial segregation and systemic disinvestment. “Deeply rooted racism in the early 1900s produced segregated neighborhoods, which influenced and perpetuated unfair home loan lending practices for several decades,” the authors write, connecting systemic underinvestment to the problem of property crime.

Higgins believes a similar pattern is at play in Dallas.

“These issues were created by design. Crime is the output, the end result of that design,” Higgins says. That’s not just speculation; Cohen says that similar analysis is possible in Dallas.

It may be then that Risk Terrain Modeling can help Dallas policymakers move toward answering the reformers' demand to “reinvest in community.” Still, Mead worries that this approach, if not managed carefully with proper oversight, could contribute to displacing minority and low-income communities in the name of economic development and public safety.

The logic is cruelly ironic. Areas in historically redlined neighborhoods, poor and crime-prone because of racially motivated disinvestment, are often identified on risk maps. These areas are then targeted for interventions that can include police patrols, code enforcement, litigation and investment or redevelopment. Depending on how it’s all done, Risk Terrain Modeling could lead to intentional reinvestments that uplift long neglected communities. Alternatively, it could result in redevelopment that prices long-time residents out of their homes.

“The reality is that communities facing gentrification are showing up on the risk maps,” Higgins says. Higgins believes the challenge is ultimately how to ensure Risk Terrain Modeling is used to benefit the communities it identifies.  “How can we center the community in the process? How can we use this as a tool to center what the community wants and needs?” Higgins asks.

Higgins is hopeful that if used correctly, Risk Terrain Modeling can help correct the cycles of poverty and violence without contributing to gentrification. “It’s innovative. I’m hopeful, but I’m watching closely,” Higgins says.

click to enlarge
Former Assistant Chief of Police David Pughes is the head of the Office of Integrated Public Safety Solutions.
Kathy Tran
Part of why Risk Terrain Modeling is so innovative is that it is a highly flexible approach for analyzing anything that shares space and time in common. “It allows you to selectively decide what to model. It can be very subjective” Pughes says.

This flexibility is a source of both strength and risk. “The cool thing and the bad thing about Risk Terrain Modeling, much like any machine learning application or geospatial application, is that you can map just about anything and come up with something interesting,” says Nick Evans, a researcher on national security and emerging technologies based in the United Kingdom.

“So the questions become: What data is going in? What relationships are being tested? What theories are informing these tests? What policy proposals are being spit out?” Evans says.

Answering these questions is central to mitigating two key risks inherent in the use of Risk Terrain Modeling: bad theories and bad data. “For example, more police equals less crime — which is just not true, criminologists have known this for decades — but Risk Terrain Modeling can’t tell you if it’s a good theory or bad theory,” Evans says.

As for data, the issue is that it is always a reflection of the history of policing in America. “In a country where the history is that police have tended to not charge whites for marijuana possession at the same rates as Blacks despite similar rates of use, the data you put in doesn’t tell you why the data is that way,” Evans says.

A recently released report funded by a National Science Foundation grant and drafted by a group of interdisciplinary academics,  “AI Ethics and Predictive Policing,” argues that predictive policing approaches are often fraught with systemic bias: “If these systems are trained on data that represent the behavior or activities of the police department itself, such as arrest records or police contacts, then it raises the possibility that recommendations become a self-fulfilling prophecy."

Cohen of the Child Poverty Action Lab is particularly sensitive to these worries. “As an independent nonprofit entity that receives no money or contracts from the city of Dallas, we can walk away at any point if we believe Risk Terrain Modeling is being utilized as a blunt instrument,” Cohen says, alluding to the history of hot-spot mapping and over-policing.

Unlike the abortive attempt in Arlington, it appears Risk Terrain Modeling is here to stay in Dallas. Cohen frames the current stage as a set of prototypes and pilots, implying that how Risk Terrain Modeling is used in Dallas will change over time. The details of how it will change are unclear. Without a formal oversight board, tracking them may prove difficult. And without adequate transparency, the concerns about self-fulfilling prophecies or gentrification are unlikely to go away.

“It’s important to lean into the complexity of risks and opportunities. But I don’t want skepticism to override progress” Cohen says. “I’m open to the idea of an oversight board.”