Workshops Tackled Big, Real-World Problems at Data for Good Exchange 2018

To understand the energy at this year’s Data for Good Exchange (D4GX) conference, held at Bloomberg’s Global Headquarters in New York on Sunday, September 16, consider this: Every single workshop was packed to capacity – and then some.

“We’re going to need a bigger boat,” said Catherine Cramer, manager of industry engagement for Columbia University’s Data Science Institute, who noted that even her organization’s funder couldn’t find a seat. “All four workshops were over capacity. Clearly, the excitement and motivation are there.”

This year’s D4GX was the first in five years to incorporate the workshop format in the program. Attendees’ desire to put their skills to use to dig into real problems was palpable. One workshop, “Addressing Community Challenges with Data-Driven Solutions,” brought together data scientists with leaders at Elmcor Youth & Adult Activities, Inc., one of the largest social services agencies serving the neighborhood of Corona, Queens.

“We’re really seeking to narrow this data divide, which is easy to say and hard to do,” said Cramer.

The goal of the workshop was to help Elmcor see how it could better use data to communicate its effectiveness – especially in the prevention of opioid abuse – to stakeholders, including program participants and current and prospective funders, as well as to design future projects and to develop a data-driven approach to their work overall. By the end of the session, about a dozen participants said they wanted to continue volunteering to work with Elmcor.

Andrew Nicklin, director of data practices at the Center for Government Excellence (GovEx) at Johns Hopkins University, led a workshop titled “An Ethics and Algorithm Toolkit for Government (and anyone else!).” In it, he walked participants through a toolkit designed to ferret out the risk of bias in an algorithm.

“Expecting a typical middle manager in government to understand how powerful the data is, or how the algorithm can help or harm, is a little unrealistic,” noted Nicklin. “How do we empower them?”

The toolkit, currently in beta, walks users through issues such as impact, historical bias, and technical bias. It then suggests strategies to manage and mitigate risk once it is better understood.

Now, Nicklin’s mission is to work with governments to help them use the toolkit, and also to encourage its adoption elsewhere, perhaps in the private sector. “We have sort of a huge future,” he says. “I’m excited to see what other people do with it.”

Trust – in data, people, and algorithms –played a role in all four of the workshops, but especially in “2020 Census Toolkit.” This workshop was designed to foster collaboration to help U.S. residents better understand the aims and importance of the U.S. Census, and most importantly, to encourage them to fill out the survey completely.

“We came up with new approaches to awareness and education campaigns,” said Erica Matsumoto, head of partnerships for NYC Media Lab, whose team led the workshop together with members of the U.S. Census Bureau’s Census Open Innovation Labs and Mark Hansen, a professor at Columbia Journalism School, where he also serves as the director of the David and Helen Gurley Brown Institute for Media Innovation.

Some of the ideas generated included encouraging student groups to look for bias in the accuracy of the census; encouraging competition between states for higher completion rates; leveraging influencers; and, of course, gamifying the whole thing.

Jake Porway also had a line out the door for the “Data for Good 2025: Building the Data Science Ecosystem We Need for Tomorrow, Today” workshop he organized. The workshop aimed to get participants to connect with other sectors they don’t normally collaborate with day-to-day and to get dreaming about what we’d need to create today to create a vision of Data for Good in 2025. Participants hashed out some of the issues around building trust, mostly through understanding what data science can and can’t do. They talked about the need to be able to talk about data science in a way everyone around the Thanksgiving table could comprehend, because, “without basic understanding, there is no trust.”

The goal of ‘data literacy,’ sometimes generates confusion, Porway said. “Just because you’re literate doesn’t mean you’re also a great writer. Sometimes ‘data literacy’ gets confused to mean that everyone has to be a data scientist.”

The other challenges in building a data ecosystem for 2025 that were discussed during the workshop were just as daunting. How can we move beyond one-off pilot projects, or “pilotitis”? How can we further democratize data? How do you create data-driven solutions, when coming up with anything novel requires expertise in so many different areas? How do we create a Data for Good community that is as diverse as the constituents it aims to represent? And what big investments do we need to make now to make sure this capacity is available to all who need it?

Porway noted that The Rockefeller Foundation has given DataKind a planning grant to kick off this conversation. He was clearly excited to get started. “This is an open invitation, people,” he said. “If you want to be part of this next big phase, reach out to us.”

Learn more about the takeaways from the workshops at Data for Good Exchange 2018 below: