Transcript

Jim Fruchterman [00:00]

Welcome to Tech Matters, a biweekly podcast about digital technology and social entrepreneurship. I’m your host, Jim Fruchterman. Over the course of this series, I’ll be talking to some amazing social change leaders about how they’re using tech to help tackle the wicked problems of the world.

We’ll also learn from them about what it means to be a tech social entrepreneur, how to build a great tech team, exit strategies, the ethical use of data, finding money of course, and finally, making sure that when you’re designing software, you’re putting people first.

Today, I’ll be interviewing Jake Porway, the founder of DataKind.

Jake Porway [00:39]
If you think about the mission of DataKind, the question that it is sort of designed to answer — in a way, it’s a giant research engine of a sort — what social impact can you have when data scientists are united with nonprofits in order to try to solve problems together?

Said differently, what could the nonprofit sector do if every nonprofit had Google’s engineering team or something like that?

Jim [1:04]
It turns out that Jake and I are wrestling with many of the same issues. For example, the question “Does every nonprofit need a software development team?” has an analogy: “Does every nonprofit need a data science team?”

And so these issues come up over and over again about how do technologists, who have highly valuable and expensive skillsets, how do we apply those skillsets to solving the social problems that people in the social sector actually have, as opposed to the ones we might imagine that they have.

Jake [01:37]
All to say, I fell in love with computers, but I was just sort of dismayed that when I got out of… I was going to get out of college, I looked around to say, well, great, now, how can I use this computing for good? Let me bring these two passions together. And I kind of looked up and saw job openings at video game companies or Wall Street or whatever. I said, okay, you know what? Let’s just keep going with this path. I went to grad school, I studied machine learning for my grad work, but actually switched over to statistics to do that. And that ended up being a really happy accident because when I left grad school I had that same question of: What do I do?Where do I bring these machine learning skills?

Most of the answers were, you know, government or big tech companies. But this was around 2010, so it’s right at the time that, you know, we’re about three years into the iPhone being out, you know, we’ve got this web 2.0 kind of in its heyday. And that was when, you know, big data was everywhere. And people were saying, hey, data is being collected by almost everything we do. And so I thought, oh my God, this is it. This is great because all of a sudden here my job options aren’t just where the data is, but in fact, with data being distributed everywhere, oh my God, like with these data science skills, as they’re being called, we could learn so much more about our world. We can start studying, you know, crop, you know, yields. You can start looking at how different social movements for, I mean, oh my God, that seemed like the possibilities were limitless.

And I was disheartened one last time to realize as I looked around, as I went to the first Strata conference, the first big data conference O ‘Reilly threw, that almost every talk I saw was about getting people to click ads. I was just like, this is… Jeff Hammerbacher got, you know, quoted saying that, you know, “The best minds of my generation are being used to get people to click ads”. And so we spoke with a number of other data scientists who said, you know, for all the attention this title is getting, couldn’t we do more? Aren’t there, you know, nonprofits that have tons of data that we could work with?

But over time, you know, I put out a little blog post about it just to say, hey, is anyone interested in this? You know, thinking that it was a dumb idea. I’m just gonna send this to like seven people that I know, because I’m sure no one else wants to spend their weekend hacking on nonprofit data. But I was very surprised to find when I got back to work the next day, my inbox for that account was just overflowing, because someone had shared it with someone else, who’d shared it with somebody. Tim O ‘Reilly had tweeted about it. Before we knew it there was this post, Data Without Borders, which is what it was originally called, “A chance for data scientists to join the call of doing good and nonprofits to help”. Which I know, it must be, you know, reliving a story of your own, I imagine, with Benetech and your work and doing tech for social good.

Jim [4:22]
No, I was alone for 10 years, so I didn’t have that instant catch fire moment, but congratulations. So you’re like a year or two into a regular job in data science and this blows up?

Jake [4:35]
Yeah, exactly. [laugh] It was kind of funny, I just started my career out of grad school and I was like, well, I guess this is what we’re doing now.

Jim [4:43]
So in those first couple of years you’re getting more and more people excited about this, you’re having a volunteer model, a data hackathon kind of model… or did it evolve more quickly to other things?

Jake [4:58]
Yes, the first format was hackathon, right. My co-founders, Drew Conway, and my other co-founder, Craig Barowski, a long childhood friend, when we co-founded we said, look, let’s just start with what’s already working. People are already going to hackathons. This was the heyday of hackathons, so let’s just tweak it. Instead of coming up with random projects, let’s have nonprofits kind of apply and we’ll filter to understand who’s got a problem or data set that looks like maybe you could possibly learn something from some data in a very short period of time.

It was incredible because we actually thought the first one was, it’s like the first pancake to throw away. We weren’t planning to start an organization, so just: let’s organize a Hackathon, see who comes, see what happens. We had the New York Civil Liberties Union — this was in New York — the New York Civil Liberties Union sitting across, being led by Cathy O’Neil, who at the time was a data scientist at D.E. Shaw, which is… we love Cathy, we know her. She’s just great at data science. Here she is working with police stop-and-frisk data, so already on the path of thinking about how are we looking at social justice and equity through data, putting some of that power back in the hands of nonprofits.

The UN Global Pulse came and brought a data set about what needs people had around the globe, and that team, who was led by one of the folks that had been the head of the data scientists at Spotify, ended up finding… building this visualization that they liked so much that they brought the team to the UN two weeks later and said, look, this is how we have to start thinking about data and computing.

These were just incredibly strong signals to us that if you had really smart, humble folks from the social sector and the technology sector working together, you could actually make some real impact really quickly. That hackathon model carried us pretty far

But then we followed a trend with DataKind’s growth of, how do you get closer and closer to longterm projects based on what people had already done. So, pretty naturally, some of the folks at the hackathon said, “Hey, I kept working on this. Our team kept getting together because a weekend wasn’t enough”. So then you’d have people going for two weekends, for three weekends. That was what led to us starting the DataCorps, which was the longterm six-to-nine month project that folks would sign up for to do 20 hours a week, 10 hours a week.

Jim [7:19]
And I think that’s… DataKind was probably best known for that sort of longer term commitment. At least that’s kind of the… how I think of DataKind and having chapters and all that. So, talk a little bit about how the organization… I mean, when did you switch from being a volunteer to actually saying, I’m going to do this?

Jake [7:37]
Oh, you know, that’s funny. It’s a good question, because I had to be dragged a little bit, partially because I just was convinced, and in some ways, maybe rightly so, that I wasn’t cut out to be an Executive Director or CEO. Like I said, I wanted to be one of the data science volunteers. That’s why I wanted this thing to exist. But also, maybe just from lack of knowledge or realization of what was possible.

So we ran a couple of these hackathons that were called Data Dives in New York. And then we said, ok, well, New York was great, but does this work elsewhere? So we tried San Francisco, great. Chicago, DC. It was after DC that a number of folks that were there, that were kind of informal advisors, and people more senior in their careers kind of looked at us, and were like, you can’t just keep doing this as volunteers. You know, we were burning the midnight oil, trying to tell our jobs why we were dead asleep on Monday, because we’d spent all weekend flying out to San Francisco, staying up for 48 hours of flying back.

And it was honestly very lucky that word had spread about the Data Dives. And Josh Greenberg over at the Sloan Foundation approached us and said, “Hey, listen, this is a real thing, you could replicate this. And if you wanted to, we could potentially support it”. And so that, I mean, it made it really easy. And we got so lucky. People work so hard to try to get grants or stuff, way more worthy than this. It’s really difficult. I don’t know, right place, right time, privilege, not sure what it was… But the doors really just got opened really easily by folks saying, hey, we see something here. And we’d like you all to support it. So that’s when Craig and I quit our jobs, went full time in 2012.

Jim [9:15]
So you evolved from hackathons, data dives to this model of a six-to-nine month engagement where you go deep with a nonprofit. So do you think data kind is mature? What does maturity look like?

Jake [9:32]
The Data Dive era was: Can you say anything meaningful in 48 hours? Can you attract the right time of tech talent? How quickly can you find a suitable data science problem in a social sector organization? It’s not trivial. And that proved itself out pretty quickly. Because like I said, you got the UN with a massive data set they’ve collected asking, even for basic analysis that could help them with policy decisions. OK, so that was true.

But then, like I said, we went to the longer term model. Because just incredible folks, I think, starting actually out of a Data Dive in DC, the DC Action for Children Project, had a number of folks who stayed on and just kept working on that over time. This was a small nonprofit that was trying to understand child wellbeing in the district. They had tons of data. People didn’t know how to navigate it. It was all in PDFs, a classic problem. And so this team converted it to a layered map, which is so funny, so easy now, but had to be done custom at the time. And they gave them all sorts of analytics and helped them realize, oh my goodness, there’s longterm capacity needed here.

The kind of next phase shift that was interesting was, well, how do you grow the data kind model? There’s so many folks who want to be on projects. But it’s not going to work if we’re all centralized in New York, where data scientists in Tokyo want to work on a project in Sri Lanka… what do I know anything about any of the context of any of that? [laugh]

But the cool thing that happened was a number of folks would write to us and say, look, I’m really interested in being involved in DataKind, but I not only want to do the projects, I want to run DataKind. The first person who did that was this fantastic guy, Duncan Ross and Fran Bennett, actually, in the UK. And they said, look, we basically could just start our own DataKind UK nonprofit. Because we know UK issues, right? And this was really early on. I think there’s a lot of lessons learned about trying to replicate that early. But they were phenomenal leaders, and they built out DataKind UK, which a lot of folks now know first over DataKind, because they were in that area and they just got a great presence.

So that was like the next chapter was, ok, so how do you not manage the projects directly anymore, but how do you train others to recruit volunteers, manage projects, find the right nonprofits? And so you saw this huge outpouring of interest, and we ended up starting chapters in Singapore, in Bangalore, we mentioned the UK already, we had some in San Francisco, DC, and the headquarters were there in New York. There was a lot more interest from other folks, but that was actually where we kind of capped it to just try to master replication at that level.

So ok, so now we’ve got all these people doing their own volunteering, to kind of train the trainers situation, but all these other people in different countries doing the work. And you know, this is, I mean, my heartstrings always get tugged seeing the projects and seeing the communal way the projects are done. And so for me, it was just so enriching to see folks tackling their own local challenges. Like I remember hearing about the UK teaming up with Open Corporates to sort of understand patterns of money moving through corporations, and ended up actually finding an analytical evidence of government corruption that actually led to a new law being passed to prevent that kind of loophole… and, you know, analytics that are a huge impact on shaping policy.

So that was the kind of spread phase. And actually, one thing that I think a lot about is how that can keep going, because that was the… it was it’s the most exciting communal effort, in my opinion, and it’s where you’re really going to get kind of equitable data science change. It’s not that like a bunch of data scientists in New York and San Francisco who are volunteering to, you know, swoop in virtually into problems around the world, but actually building up ways for the local people in local areas to solve their own problems. It’s just hard to get funded and it’s hard to do quality control, which we could talk about.

But then the very last thing (sorry I’m talking a lot about this), so if that was kind of the global volunteer scale phase, the last phase or the current phase, I should say — last for me, because I departed about a year and a half ago [note: at time of recording], was looking at the kind of issue area or systems level, because DataKind’s goal was never to run the most volunteer projects. Like I said, it’s sort of to say, what problems could you solve? How do you keep pushing the envelope on what problems you can solve if you have this tech capacity? And after, you know, hundreds of these projects are coming in from around the world, you quickly see people are starting to solve the same types of problem. I can’t think of how many people in the 2014s tackled trying to do credit scoring on non-traditional data. And so you start to ask the question, well, if this is such a common problem, and you’ve now had six pilots in different contexts, can you learn anything that might help you figure out what is a more general solution to this problem? Or why it can’t be generalized? Why is it so context specific?

And so we started a program called Impact Practices where we would intentionally focus on an issue area, say community health workers who are recently… they’re folks who go toward people who are providing healthcare in areas where there’s limited health infrastructure. And there’s a question about, for all these organizations using digital tools in community health work, how might data science and machine learning help with health outcomes there? And that’s a big, thorny question. There are a ton of problems that have nothing to do with data there. There are tons of problems that are just from a purely equity framework question, even before you get any tech involved. With all those caveats, the question we sort of remained with, was there something in there that enough of these organizations had in common that it could be a bigger lift solution that should be solved? And teams worked together and found all sorts of automated data cleaning techniques that could serve many organizations. They found why it was actually too difficult with the current data landscape to predict who was gonna need more urgent healthcare sooner. That was a project that often got put up for funding, because it’s easy to think, oh, if you have all this data about patients, maybe you could predict who needs help before they even know it and then send the community health worker there sooner. Great thought when it’s described in that sentence, not very easy to do practically when you actually look at what data is available and what the processes are. So anyway, that was the kind of, I thought an interesting development in the phase of DataKind in thinking about how do you solve that type of problem.

Jim [16:24]
I mean, did you guys think you were going to build products that people could use throughout the sector, the specialized tool for doing X or Y?

Jake [16:33]
The thing that we’ve gotten good at that, I think, you have to really cue to as an organization is harnessing latent volunteer capacity. That’s not the kind of capacity you want to use for software design and maintenance and upkeep. It might be… but it is pretty good for a kind of R&D effort, you know, a bounded engagement to learn… to basically do feasibility testing. And I hope anyone from DataKind listening to this doesn’t feel this is a knock at all on DataKind. To me, I think one of the things the social sector lacks the most is that kind of Proof of Concept generating ability when it comes to data science and computing. People can either come up with imaginative ideas, but then not be able to deliver on them, or they can take Proofs of Concept and build them out into, you know, sustainable systems.

Jake [17:29]
That’s, I mean, it’s not easy to do, neither of those is necessarily easy to do. But it’s… data science is super unique because the solutions you’re building, if you’re building a predictive model or an algorithm, often depend on your partner’s data. And there’s no easier quick way to just look at someone’s data and say, “Oh, I know what I can do with that”. You have to actually put in the… you have to do fairly complex analytics thoughtfully to even know if there’s a problem to be solved there. And so that’s just… that discovery process, I think, is so critical and not, you know, it’s fairly underfunded and maybe not appreciated.

Jake [18:03]
So all to say, did I think we make products? I didn’t think DataKind would in-house have a development team, but I thought we could try to de-risk the space for others to say, look what is possible, with this data under these conditions you can predict this in the hopes that we can either partner with other folks or have them pick it up.

Jim [18:20]
But I think you’re describing an arc to deeper and deeper engagement over time. And eventually, people need to bring this in-house, have this capability. And you’re describing DataKind as the R&D lab that helps you go, “Oh, maybe we should actually bring this capacity in-house because we’ve seen what it can do”.

Jake [18:41]
I love where you’re going with this because… I love your thoughts on this knowing what you do at Tech Matters because I think you’re tackling this problem really head on and solving it in a lot of ways. I would say the evolution… if I could map out the problem that I would like DataKind to solve next, or anyone to solve next… where I would go in this trajectory is, like you said, you’ve gotten this deeper engagement, you’ve shown now how to move from individual problems to kind of problems that many people can solve, and what that, I would hope, would lead to would be the answer to the question: How does the social sector generate and maintain data science and machine learning AI solutions for its own problems?

Right now, because there’s a lot of friction to getting the tech capacity, the right tech capacity, finding the right problem, knowing how to build it… because there’s friction there, DataKind can kind of reduce those barriers to entry. Well, it’s one way to do it… I don’t think it should be the long way to do it, that you just have this giant nonprofit with a bunch of volunteers. But then what is the right model? Does every nonprofit need a machine learning team? Probably not. Will all machine learning talent for the social sector actually live in the for-profit sector and then just lend its time? I think that’s also probably not the answer. But then what is it?

I think that’s one of the research questions. I don’t have a strong hypothesis for it because I think we need more of these… to kind of build demand with more of these proof points where we say, “Hey, the community health workers, with this algorithm, dramatically increased health outcomes. And everyone’s using it”. And with the hopes that then people would perk up and go, “Oh, we want more of those types of solutions”, not the one-off apps, not one nonprofit getting a little better at data maturity, but that kind of thing really made a difference.

Now, how do we make more of those? And that’s… And who does it, and what institutions do you need to make that happen longterm?

Jim [20:36]
Well I think that, you know, most organizations operate on outsourced technology platforms to do a lot of their fundamental things, human resources, accounting, and so on. And so, I think we’re talking about upgrading the capacity of social good organizations to make an impact by having better tech tools. And in this era, the idea that data might not be that important is ridiculous. [laugh] You know, and so… and the for-profit sector, you know, is 10 or 15 years ahead of the nonprofit sector, you know, in these things.

So in a lot of ways, we’re not talking about figuring out how to do things, or we’re not necessarily even building the tools that you’d use. I mean, there’s a lot of off-the-shelf tools that can solve, you know, many of the problems that we… but if you don’t know about those off-the-shelf tools, then you’re not using them. You don’t know how to use them if you don’t know how to… to know when they’re working or not, all that kind of stuff. And so I think, and I think this is kind of one of the fun sort of field-level questions, is how do we raise the game of the social sector so that, you know, ideally they can get twice as much done with the same amount of money or the same number of people.

So, if you think about that, how does the social sector, you know, solve problems better, figure out how to have better health outcomes, better economic outcomes, better, you know, educational outcomes, you know, whatever your goal. What are the barriers to making that happen? You know, and how do we get rid of those barriers or, what bridges do we need to build… so, what are your thoughts on that one? Cause I know that you have been wrestling with these kinds of questions.

Jake [22:21]
I think that’s such a great question and there are, I’d say, two or three major challenges to the social sector adopting data and AI technologies, or becoming as data-driven and AI-driven as they can be.

The first one I would say that actually is missing is lack of a clear vision. So you’ll hear phrases like, “We need to be more data-driven”, “We need an AI-enabled social sector”. If you start to pick a little bit and ask yourself, well, what would that look like in the long run? What would it look like if every time an SDG indicator fell below a certain mark, where technology was useful to bring it back up, it was deployed, right? Like, it’s a strange framing, but if there was some need for impact, the technology could solve it and you made it. Well, it raises questions like we were asking before, who would solve that? Would every nonprofit be in charge of building its own machine learning solutions to do so? Or would there be kind of centralized organizations that did that?

When we talk about what’s most important, I said machine learning, that usually implies algorithmic efficiency, implying that “we could do it” — we just need a computer to help us do it faster and cheaper. But of course, as we know, the applications of data and computing also apply to, say, modeling and testing aspects of our world to understand what systems are working. So there’s a research component to the uses of data and computing. And so, if we’re not specific on what it means to be data-driven or data-enabled, then it’s not clear what vision of the future we’re headed towards.

I’ll hear often folks say, we need to be more data-mature and it’ll be treated like it’s this binary, like you’re either not good at data or you’re really good at data. And being, quote, “Really good at data” usually means you have some algorithms or some programming or custom computing. But I would always ask: The organization, the nonprofit that’s building a piece of software for community health workers, probably has a very different data and tech team than, say, the organization that is an advocacy organization, that’s to convince lawmakers; do they both have the same kind of team? No way. In the same way that athletes have different specialties, I think the first team needs software designers and user experience to help drive their theory of change. Whereas advocacy folks might need to use more data visualization or data for persuasion, because at the end of the day, you’re trying to get somebody to believe in what you care about. That’s not an efficiency problem.

So, kind of mixing a couple of ideas here, but that vision clarity of what we want, I think, is number one. And it bleeds into number two about what is the use of data and computing that is most useful for the problem you’re trying to solve. And I think there’s a huge language problem where we say “data”, we say “AI”, we imply that we know what that means, but people vary based on what they’re talking about in terms of the data they’re talking about, the outcomes they care about. So I think that’s the other thing when you talk about, what is it that could help people get to this next level, it’s clarity of language, what do we mean by outcomes from data and computing?

The third thing I’d say is we have to get out of thinking about data science and AI as modeled by the for-profit sector in the nonprofit sector. I see a lot of implicit copy and paste. When I hear funders talk about it, they say, “How could we get scale?” And I think what they mean is they’re picturing an app like Uber or Lyft, where you hear about these companies, they created a ride share app and before you know it, everyone had it on their phone. And so they’ll look at the voter registration tool or crop yield app that somebody built and go, “How do we scale that? How do we get that to everybody?”

But that’s a scale that scales a very specific type of for-profit problem, which is efficiency gains that everyone has in common. You know, Lyft and Uber scales because you and I need rides places and people want to give rides and that’s a transaction that a computer can speed up for us. But social sector problems can be totally different than that. You don’t always have just that transactional thing that every individual has. You might have complex challenges like even on the efficiency front, like managing food supplies for food banks. Well, even if every food bank itself is using that, there’s a kind of meta problem of: How are all the food banks coordinating? Because if you’re really trying to solve the problem of feeding people, you don’t just do that by feeding each person one-to-one. You have to think about the bigger picture. So the simple kind of like, “take an app and scale it model” may not work.

Also, I think you see a lot of the kind of pushback to technology in the social sector because it imitates a lot of the extractive nature of for-profit tech. Implicitly, I don’t think it’s, no one’s trying to do anything bad. It’s just that’s how we’ve seen tech used. So we’ll collect constituents’ data and we’ll think about how we can analyze that data so we can keep making money as a nonprofit vs. the power we have in the social sector for foundations to say, no, we’re not going to fund models like that. The data needs to belong to the constituents, for example.

Jim [27:44]
You’re surfacing a lot of really critical issues here, right? So I do spend some time talking to nonprofit founders who’ve been told by their donors, often venture capitalists, that, you know, how are you going to get a lot of eyeballs? How are you going to get a lot of data? How are you going to monetize that? Don’t you want yourself not to be sustainable? Don’t you want to essentially monetize the data of the poor or the suffering? And actually the right answer is… no!

Jake [28:11]
No, absolutely not! [laugh]

Jim [28:15]
And so I think, you’ve also identified another problem, which was, it’s just not monetization. Uber and Lyft add value to us as riders. We’re happier with the service we get on the average than we got from taxis. But they, you know, they introduced a new thing and they become worth many billions of dollars. And, you know, are the drivers happy with this? It’s a mixed bag, right? And so, and I think you described that basically as the extractive nature of many of these models is, you know, we can get rich by making poor people poorer and making rich people happier. Okay, the private world’s pretty good at that. That’s not what the social sector says.

And then I think the prior point you were making is about, you can’t do data or AI or data maturity sort of independently of your social goals. But I think we’re taking a step back of how to upgrade the entire social sector’s ability to do its work, and we think that being better at data should help us figure out that this doesn’t work and this does work. Or we could do three times more work for the same amount of money or whatever. So there are these sort of definitions of social good that the technology, the data, should be able to support. But it’s a big challenge because nonprofits don’t have a lot of money. And data scientists generally can make a lot of money. Given our economy these days. So they’re unobtainium.

Jake [29:51]
Yeah, exactly. [laugh] I think for all of the big kind of philosophical challenges that we talk about in moving the social sector forward and figuring out what a vision of a data-driven social sector looks like, it pales in comparison to the practical challenges that you’re referring to where even access to the technology or the data, technologists, is super restricted. And so we’re behind in that regard.

Jim [30:20]
So what, so let’s, let’s talk about the human capital side. I mean, and this is not an uncommon problem, right? There’s, there are these people who have a high skillset. Society could benefit from it. The for-profit world will pay them far more on the average. And yet, some people decide to do social good in spite of the fact that they could be making it two, three, four X more. Um, so, I mean, what do you think the mix is? Do you think it’s, you know, some expert data scientists through a group like DataKind coming in here and trying to raise the game of a field while the people inside that field get better and better, or regular full stack engineers get better at this or… I don’t know what does it look like in five years?

Jake [31:04]
So I think we’re at an interesting time because there is still excitement around data science and AI; for as much there’s also a lot of concern. And so I don’t know what the final model will look like, but we do have some examples of other professions that do this. So lawyers have a robust pro bono practice. So, maybe, the pro bono AI engineer data science model could be… viable. But that’s probably just still handling one-to-one cases. And another thing to think about is: Are there kind of in the same way there are low-cost clinics or institutions or consultancies in other fields that work for the nonprofit sector, this kind of big machine learning consultancy for social good? I think there is a version of that model that’s not just relying on volunteers and the people’s goodwill. To give people good jobs and also provide good outcomes and talent to the sector, I think there’s a world for that.

Jake [32:08]
I think either way, though, what I would hope we could use this energy around data science and AI for is to kind of give a shout at the funders to say: What you fund is what will exist. And if you want to see more technologies in the social sector, don’t encourage more employee volunteering programs. Fund a center of really good machine learning folks who don’t have to make a difficult economic decision.

I know that that is not going to be a popular stance. And every industry or every issue area ever has complained, “Well, there’s not enough funding for this”. And so it’s a perpetual problem.

Jim [32:48]
Well, I mean, I often joke, I think the way that funders as a group approach these issues causes people to bury their data science capabilities a level below and sell the social impact that they’re doing. So, I mean, I joke that an awful lot of the Skoll award winners, if you peel back the hood, are software and data plays at their core, right?

Jake [33:16]
Yeah. That’s a really good point.

Jim [33:16]
Kiva, it’s a microcredit lender. But what are all banks and lenders? They’re software companies that are analyzing data, right? And they have a solid sideline in handling money, right? So I find it quite interesting to see how the way our field works, we play down the data science capabilities of these organizations even though a lot of their promised social impacts are there because they have some sparks there. So it would be great though, if we were, like, more upfront, like in the law field.Tthere’s plenty of public interest law firms, how many public interest data science nonprofits are there? Not very many.

Jake [33:59]
Right. It’s true. Yeah, and and you know because you mentioned public interest, of course I have to nod to the Ford Foundation’s push for public interest technologists. I think they founded public interest law or started that language back in the 60s. So it’s sort of natural that they would push public interest technology as well. Yeah, I completely agree there, whether it’s through some US Service Corps or through a kind of branch of the profession, I do think you’re right, public interest technology, public interest — you know, within that data science and machine learning specifically — is gonna be critical for us getting there.

Jim [34:35]
Well, this has been a lot of fun, wide-ranging conversations. I want to make sure to give you a chance if you have any sort of final thoughts, like if people wanted to follow in your shoes and start a Data for Good organization or…

Jake [34:50]
Oh, great question. If you’re someone listening to this and you want to go use your technology skills for good, sign up for DataKind, sign up for Code for America, sign up for whatever your local volunteer tech organization is, just do it. Even if you’re scared to do it, do it. These communities are incredibly kind and generous and you will learn so, so much. There’s so many folks who I know want to do this kind of work. Try to find a way to get out and just try working on it because the world needs you to do it.

And I think if you were to start an organization around doing Tech for Good, I think, I mean, now more than ever, the question I would have is, who are you building it with? Obviously, early on, one of the first calls we had, one of my favorite all time calls that DataKind started was with Lucy Bernholtz at Stanford. So we said, we’re just technologists. We don’t know anything about nonprofits. And she went on a beautiful rant about the problems that nonprofits face and how you don’t want to burden them with these basically white guys from San Francisco coming in and going, let me look at your data because this is going to be really fun. It was incredibly informative in how we tried to keep kind of a humble lens of what we did, but even we didn’t go far enough. And I would say now, if you’re building, now that we have thankfully started to reach an era where people are more aware of the ethical implications of technology and data, when it maybe shouldn’t be used and the problems that could occur when someone builds it out of context, I would ask more, if you’re going to do this, who are you building it for? And how are you making sure that the folks who are impacted most by the technology are deeply, deeply involved. And I don’t mean involved as like giving feedback, but have power in deciding how things go.

Jim [36:33]
Well, thank you very much, Jake Porway, for joining the Tech Matters Podcast!

Jake [36:36]
Thanks so much, Jim, a total pleasure!

Jim [36:39]
Thanks for listening to this episode. I’m your host, Jim Fruchterman, and if you have any questions or comments, feel free to send them to us, [email protected], and spread the word. If you enjoyed this, let somebody else know, let 10 people know, and feel free to leave us a rating on your favorite platform!