Page 45 - GCN, August/September 2017

where, when we send our rodent con- trol team out, they’re actually finding and being able to treat rat burrows,” he said. “We’re then going to try and predict those locations using environ- mental variables that are known to be predictive of rat ecology as well as information about the geography of the city, such as where we have lots of restaurants, where we have lots of construction, where people are most concentrated.”
Although Casey said he could have written such a model before now using public data, significant context is impor- tant to understand before writing the code. In his time with the rodent control team, he learned why the spreadsheet looks the way it does. He also learned why less urban areas have fewer 311 rodent calls and which environmental factors lead to rat infestations.
And conversations he had with a ro- dentologist helped him determine that his unit of analysis should be a census
block because rats don’t like crossing streets or natural barriers such as riv- ers, Casey said.
The lab is currently working on near- ly 20 initiatives, including studying the effectiveness of body cameras, seeing if predictive modeling can improve hous- ing inspections and assisting in efforts to reform the city’s criminal code.
TRANSPARENCY AND
DATA ETHICS
The Lab @ DC team is using a variety of methods to analyze its data, includ- ing R, Python and JupyterHub, an in- terface for scripting in Python that runs on a Microsoft Azure server. “As a team, we’re trying to develop our capacity to work in Python,” Casey said.
The lab borrowed most of its data policy from the city’s existing policy, but data ethics are also a major con- sideration because of how harmful a model can be if it isn’t written cor- rectly, he added.
“Machine learning models can per- petuate historic inequalities that are part of the society we live in,” Casey said, “and we have to be constantly vig- ilant about how to use the data we have to serve our populations and avoid that as a possibility.”
Yokum added that transparency about the data sources and the model- ing decisions also helps ensure ethical data use and the public trust.
City Administrator Rashad Young said in a statement that the Lab @ DC will directly inform the decisions D.C. officials make.
“You can’t manage what you can’t measure, and by using the scientific method, we are getting the best pos- sible measurements to inform how we manage the city,” Young said. “That means we are learning from the evi- dence that exists in the world while taking the next step of generating our own evidence so that we can know what works in the D.C. context.” •
Some of the other projects the Lab @ DC is working on:
• Analyzing crime trends in D.C.
• Exploring how the criminal code is used
• Evaluating the DC Crime Gun Intelligence Center
• Improving police recruitment
• Testing a flexible rent subsidy
• Creating a consolidated survey platform
• Improving the Department of Consumer and Regulatory Affairs’ outreach and education
• Evaluating the use of a 911 nurse triage line
GCN AUGUST/SEPTEMBER 2017 • GCN.COM 45