David Corliss

Jake Porway

Jake Porway is one of the most outstanding figures in Data for Good. Early in his career, he worked in machine learning for image recognition and as a data scientist for The New York Times while also leading volunteer projects. In 2011, he founded DataKind and serves as its executive director.

What do you think the state of the art is? A lot is happening in America right now, and I’d like to get your thoughts on what the situation on the ground is and how data can help.

It’s really been interesting to see what has changed and what hasn’t the last 10 years. We have a space of statisticians and machine learning folks looking for good—ready-made improvement. I think the case has been made, though, for data science and machine learning to be applied to social impact. There’s obviously a ton of questions about how to do that capably and ethically, but I think we’ve done a good job of helping folks on a local scale.

Our goal isn’t to do the most volunteer projects in the world, but rather have the greatest impact. The thing I want to see is a world in which our oceans, water, and air are clear—where there’s less violence and inequity. So, I think there’s a gap right now in the Data for Good movement between helping, say, five NGOs working on singular interventions versus saying we have built some kind of data-driven solution that overall advances our social sector. I think one of the challenges is you don’t have enough people who can articulate where those field-level problems are and where data science can be most helpful. At DataKind, we launched what we call “impact practices” to address large-scale, sector-wide issues.

There’s a lot of intersectionality—interactions between different issues. We’re looking at both COVID-19 and the election year, a lot of protests around police violence, and just generally violence toward persons of color.

Yes, this year we’ve seen a lot of pressing issues in which Data for Good can make a difference. DataKind’s Frontline Health Impact Practice is addressing health care delivery around the globe, and since the COVID-19 outbreak, we’ve leveraged data science solutions for relief efforts.

Our Economic Resilience Impact Practice is focusing on financial inclusion, keeping people above the poverty line and helping them sustain economic shocks like COVID-19. There is plenty of opportunity for our data.

It’s interesting that you had already flagged that as an area you want to focus on and then, when the COVID crisis came up, it was made a lot worse—like a toxic environment resulting from people who don’t have good insurance, or commercial operations living on the edge. Almost as if COVID-19 is creating the final result for something many years in the making.

This is where I think there’s a sobering commentary on how much data science and technology have a role to play. One of the big conversations in our space right now is the ethics of data use, your model, and, of course, the ethics of AI predictive models that are used. Our craft has largely been used to make stuff faster and cheaper. However, we have to think about these core issues of social justice and addressing our inequitable systems, all of which comes down to human issues, human compassion, and thoughtful consideration.

COVID-19, this isn’t just a medical problem. How do we fix the system that’s driving higher rates before the next one happens?

Having a data solution isn’t just about having an analytic or predictive model. We can look at homelessness and its drivers, or we can look at economic resilience. I can come up with a predictive model like FIFO or something, but what do we do with that? We need to have a call for change and drive that change. We need to look at the big picture and the underlying root factors. It’s a long journey to get there.

There’s a dearth of people who are able to manage analytic projects. If I could take on one thing, it wouldn’t be a better model; it would be a better training program for project leaders.

Allen Downey at Olin College of Engineering said there are three phases of the design process: finding the right model of decision; building the model prototypes; and managing that process. And in school, we all focus there. Now, a lot of data scientists are asking what to do with questions like, “I’d love to hack into COVID-19 data. How can I help?” or “What can we do with data for racial justice?” Instead of focusing on the middle, though, we need to focus on defining the beginning—what is the right problem to solve?

What kind of opportunities do you see today?

Where there’s enough commonality between different Data for Good projects, we would love a tool—something close to off-the-shelf that could solve prevailing problems. Wouldn’t that be great? But we’ve got to be sober about where data science can actually help. A lot of the issues we’re facing in the world today are because of systemic inequity. And that is something that’s deeply human, stemming from greed and selfishness.

The silver lining is that data science and AI just supercharge the complex of the system that exists, but they aren’t necessarily the root cause. Data science doesn’t inherently have its own ethics, but rather reflects the motivations and biases of the people and entities employing it. We have an opportunity to develop digital solutions ethically that make meaningful impacts on our most pressing issues. And there are opportunities for companies to get involved here. Every dollar counts and can translate into saving lives. Engaging the private sector can help change outcomes on a global and sector-wide scale.