Wander up to the second floor of Stanford’s Huang building and you’re likely to find grad students, data scientists, engineers and journalists systematically planning to overhaul government policies for criminal justice, education, voting rights and more.
The lab’s not-so-simple mission? To use technology to turn academic research findings about social problems into real-life changes to public policy.
“I see the Computational Policy Lab as a natural extension of the work that my students and I have been doing for the last several years,” said Goel.
He came to MS&E in 2015 motivated to delve into social issues using data science and to train the next generation of data analysts. Goel has done a lot of both, working closely with graduate and undergraduate students to build algorithms for data-driven research about issues such as racial bias in policing, political polarization and decision-making practices in the criminal courts.
But Goel identified a missing component for that research to make an impact in society. While understanding a problem is the first step in fixing it, that still leaves the second step, which is actually fixing the problem.
For instance, in one of the lab’s projects, data analysis of bloated jail populations suggests that a large percentage of pre-trial defendants could be released immediately without compromising public safety. But how can cities get that done? Which defendants can be released today? And how can they ensure that the judgements are applied fairly? They need tools and procedures: they need the solutions to the problems the data analysis uncovered.
“We’re working to solve these problems,” said Goel. “Data scientists and engineers sit with us and work in close collaboration with the grad students and with me, and together we design solutions to these issues.”
Seven new staff members have joined the Computational Policy Lab this year, including three engineers, a fresh crop of data scientists, and a second data journalist.
Dan Jenson is one of the newly-hired engineers. He’s currently working on a 53-city expansion of the Stanford Open Policing Project database. According to Jenson, creating technology-based tools to analyze government data, like searchable databases and dashboards, requires expertise beyond traditional academic research. He said, “As an engineer, you have a completely different mindset. You need to do a lot of scaffolding to build systems to scale and a lot of research isn’t built like that.”
Understanding scale comes easily to Jenson, who spent five years at Facebook prior to joining the lab. He talks about “code smell,” by which he means that he can tell if computer code has been solidly written the way the rest of us can tell if milk is fresh.
Data scientist Elan Dagenais collaborates regularly with another of the lab's engineers, Joe Nudell. "Sharad has created this special space in the academic world where you can get good work done faster by drawing in resources from a lot of different people," said Dagenais.
In addition to writing software and creating data tools, engineers fulfill another critical need for the lab, which is dealing with the extraction of data from the spectacularly outdated computer systems in local governments around the country.
“Police departments have some of the worst data systems that I have ever seen. I’ve had to write dozens of programs to convert archaic database formats into something that we can even consume…It’s like going back in time 30 years, but using modern tools,” said Jenson.
Among the data scientists joining the lab this year is Alex Chohlas-Wood. He brings critical expertise from a previous job designing analytics solutions for the NYPD. During his work there, he created a Netflix-style recommendation algorithm for crime-solving, allowing officers to identify crimes in other New York City precincts with similarities to their own cases.
For the lab, he’s helping the city of St. Louis reduce their local jail population by 40 percent in the next five years. This herculean effort will be aided by the creation of specialized data tools.
Chohlas-Wood believes that cities need to be able respond to real-time data. They should, for instance, be able to use their computer systems to see trends in incarceration for certain zip codes, or to determine the number of people currently in jail for parole violations. He said, “Without that capability, it hampers the ability of a city to understand what’s going on and craft effective policy.”
Another new hire, Elan Dagenais holds both a law degree and a PhD in economics from Stanford, making her critically important for the lab’s criminal justice work. One of her current projects is a collaborative effort with the city of San Francisco to study the effects of sentencing enhancements (like Three Strikes) on prison time. She gives kudos to the city for providing the lab with the raw data. “Transparency is so important if you’re trying to understand how your system is functioning and how it might be impacting people,” said Dagenais.
Stanford Journalism Professor Cheryl Phillips co-founded the Open Policing Project with Goel in 2017. With the Computational Policy Lab, she continues her efforts. This year, she added a teammate to the lab, New York City investigative journalist Phoebe Leila Barghouty (Stanford Journalism MA ’15).
In support of the lab’s many projects, Barghouty wields a duo of super-powers: a thick skin and incredibly strong persistence.
While every MS&E grad student wishes that government data files would fall magically from the sky, the reality is quite different. Recently, Barghouty helped file 50 separate Freedom of Information Act (FOIA) requests in the state of Texas to extract data for the lab about how many people go to jail because of their inability to pay municipal fines. (And yes, debtors’ prisons are illegal in this country.)
When local governments get these requests from the lab, it triggers a barrage of calls to Barghouty’s cell phone. She said, “We’ll get calls to clarify what we’re asking for, or they’ll try to intimidate me and say that they’re not required to answer these things… People try to negotiate. People try to convince you that what you’re looking for doesn’t exist…But the number one response I get is that they’ve never had to pull this information before and so they don’t know how to do it.” But she keeps going.
When asked how she sees her role in the lab, she said, “I am the liaison between these really gifted engineers and scientists and these very stubborn local governments.”
Another important responsibility of Barghouty and Phillips is to help educate other journalists about how to use the lab’s public databases. The recent expansion of the Open Policing Project into 53 cities provides a treasure trove of police traffic stops from Philadelphia, San Diego, Baltimore and Tampa, to name a few.
The database includes an abstract for each city that acts like a CliffsNotes guide for getting data from that local government. Which department, for instance, holds sentencing records? Which statute do you need to cite when requesting information from the municipal court?
“Other journalists won’t have to go down the rabbit hole that I’ve had to go down several times,” said Barghouty.
When asked how the lab comes up with their projects, Professor Goel said that ideas come from everywhere: lab team members, current students, previous students, other Stanford professors, and from the policy makers themselves. Anyone can propose a problem, and the group will turn it around like a Rubik’s cube to decide whether or not to pursue the lead.
Marissa Gerchick (Stanford Mathematical and Computational Science ‘20) is working on a project for the lab about pre-trial risk assessments and California Senate Bill 10, which was recently approved by the legislature. She said, “Professor Goel does a great job making the lab’s resources and team open to students.”
Goel has even involved students from his MS&E 330: Law, Order & Algorithms class (that he co-teaches with Phillips) in some of the data analysis.
While the lab’s early work has focused on criminal justice reform, it has expanded into other fields. For instance, lab fellow Josh Grossman and his team are currently beta-testing a chat-bot with high school math students to improve skill acquisition. PhD candidate Camelia Simoiu is analyzing ransomware attacks. And Professor Goel and PhD candidate Sam Corbett-Davies recently published a paper about designing fair algorithms for decision-making. That work will likely impact current and future projects.
And so the work continues to expand, bringing in a broader community of researchers who are hoping to answer some of the toughest societal questions—one algorithm at a time.
For more information about the Stanford Computational Policy Lab or to share a lead, please contact the lab.