August 31, 2015
UW students put data science skills to use for social good
The Data Science for Social Good team that worked on the family homelessness project.Craig Young / University of Washington
They could easily spend their days poring over statistical methods for a genetic study or sorting through data about consumer behavior on the other side of the globe.
But this summer, data scientists at the University of Washington’s eScience Institute took a break from their typical work helping researchers and professors to incorporate cutting-edge technologies and data-based methods into their academic pursuits. Instead, they harnessed their expertise to address pressing urban issues closer to home.
In June, the institute launched the UW’s Data Science for Social Good program, an initiative that paired data scientists with students and local nonprofit and government partners. These interdisciplinary teams worked on projects to reduce family homelessness, improve paratransit bus service, foster community well-being and map better sidewalk routes for people with mobility challenges.
The initiative, modeled after similar programs at the University of Chicago and Georgia Tech, fits with the eScience Institute’s mission to advance data-driven science in all fields, said Bill Howe, the institute’s associate director.
“Interdisciplinary is part of our brand, and the social good aspect is a powerful extension of that,” he said. “People want to have an impact. This was a chance for students to apply their skills to projects that have some relevance.”
One student team worked with the Bill & Melinda Gates Foundation and the Seattle-based nonprofit Building Changes to determine the best combination of services and programs to lead homeless families to permanent housing. The two organizations are working with King, Pierce and Snohomish counties on a multiyear endeavor to halve family homelessness in the region by 2020.
Program fellows attended tutorials on a variety of topics.Craig Young / University of Washington
One major, data-centered challenge to these efforts is keeping track of homeless families as they move through government and nonprofit services. The UW team took on the task of linking records representing homeless families across different services in the three counties, a challenging undertaking since counties don’t necessarily track or define households the same way.
“The first problem we encountered in this pipeline was defining households,” UW graduate student and team member Chris Suberlak said during a recent presentation of the team’s findings. “It would be ideal if the household ID that was provided was consistent. In at least one county, it wasn’t.”
After grouping individuals together into the correct family units, the team developed criteria for defining a single instance of homelessness. They aggregated information about the programs and services families used during that single episode — for example, staying in an emergency shelter and then moving into transitional housing.
The team also developed an interactive diagram showing the different paths families took through the system of government and charitable services. They created a Sankey diagram, typically used to show the flow of energy or costs between two points, to help providers identify programs and trajectories that may help homeless families secure permanent housing.
Gates Foundation spokesperson Anne Martens said the team’s work has already made a difference, since it analyzed data at a level beyond what the counties are equipped to handle. After looking at the Sankey diagram, she said, a county worker recognized that something was amiss in the data from a provider and was able to fix the problem.
“It’s already been a boon to the counties’ decision-making,” Martens said. “They would not have had the time or the capacity to do this without the support of the data scientists.”
Teams worked out of the eScience Institute’s UW headquarters in the Washington Research Foundation Data Science Studio.Anissa Tanweer / University of Washington
Another team worked with King County’s paratransit program, which provides door-to-door bus service for people with disabilities. The team analyzed the highest-cost rides, developed more precise usage predictions and created a web-based tool that immediately locates alternate buses when a bus breaks down.
The third team mined data to develop a community well-being framework, measuring indicators such as neighborhood diversity, socioeconomic status and places to connect. They created an interactive online map that shows well-being indicators for Seattle’s neighborhoods.
The fourth team built on the success of an award-winning app developed last spring. The app shows maps of Seattle’s sidewalks, highlighting obstacles and elevation changes for people with mobility challenges, particularly those in wheelchairs. The data science team used computational geometry and routing algorithms to construct a graph that connected the city’s fractured network of sidewalks across street crossings, then created an algorithm for the app to map customized, accessible sidewalk and crosswalk-based routes around the city that avoided curbs, construction sites and steep grades.
The four projects were chosen from among 11 proposals submitted. More than 140 students from UW departments ranging from political science to mechanical engineering applied to participate. Sixteen undergraduate and graduate students were selected, along with six high school students from the UW’s Alliances for Learning and Vision for Underrepresented Americans program.
The 10-week internship also included tutorials for students on programming languages, presentations from local tech companies and readings on various topics. About half of the interns came in with little prior programming experience, said Micaela Parker, an eScience Institute program manager.
“They learned enough to be able to advance these projects and make them happen,” she said. “That’s pretty phenomenal.”
Ed Lazowska, the Bill & Melinda Gates Chair in Computer Science & Engineering, said the initiative demonstrates the utility of data science in tackling a host of societal challenges that students are eager to work on.
“I think people are energized by the ability to work on something that is both technically challenging and makes the world a better place,” he said. “That’s what Data Science for Social Good is about. It’s technically challenging and it leaves the world — or in our case, the city — a better place.”
Tag(s): Bill Howe • Data Science for Social Good • eScience Institute
