Engaging with governments on AI for good

At this year’s Government & Public Sector R Conference (or R|Gov) our very own Daniel Whitenack moderated a panel on how AI practitioners can engage with governments on AI for good projects. That discussion is being republished in this episode for all our listeners to enjoy! The panelists were Danya Murali from Arcadia Power and Emily Martinez from the NYC Department of Health and Mental Hygiene. Danya and Emily gave some great perspectives on sources of government data, ethical uses of data, and privacy. Featuring: Danya Murali – GitHub https://github.com/dmurali1, LinkedIn https://www.linkedin.com/in/danya-murali-947153150, X https://x.com/datadanyaEmily Martinez – X https://x.com/emmartinez47Daniel Whitenack – Website https://www.datadan.io/, GitHub https://github.com/dwhitena, X https://x.com/dwhitenaShow Notes: The Government & Public Sector R Conference, or R|Gov for short, is returning in Fall of 2021 and the R Conference New York is returning in 2021 as well and you can learn more at rstats.ai http://rstats.ai/, twitter.com/rstatsnyc https://twitter.com/rstatsnyc and twitter.com/rstatsdc https://twitter.com/rstatsdc. Upcoming Events: Register for upcoming webinars here https://practicalai.fm/webinars!

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  One thing that I strongly advocate for is to be doing race conscious data analysis instead of racial blind analysis. So using things like zip codes can very, very easily be a proxy to race and socioeconomic status, but that doesn't necessarily mean that you should not look at those things. Like just covering your eyes and being like, I'm not gonna look at race. I'm not gonna touch race is not the, not the way to do it. To make sure that you are being equitable and you are thinking critically about how your data analysis affects different people. You need to really be thinking about who those people are, where they come from, what their needs are. And to do that, you should be looking at things like race and like zip codes and the relationship between those things when you're doing your analysis.

• 0:42Speaker 01

  Bane with for change log is provided by Fastly. Learn more at fastly.com. Our feature flags are powered by Launch Darkly. Check them out at launchdarkly.com. And we're hosted on Lino cloud servers. Get a hundred dollars in hosting credit at lino.com/changelog. This episode is brought to you by Digital Ocean, droplets, managed Kubernetes, managed databases, spaces, object storage, volume block storage, advanced networking like virtual private clouds and cloud firewalls, developer tooling to robust API and CLI to make sure you can interact with your infrastructure the way you want to. Digital Ocean is designed for developers and built for

• 1:21Speaker 01

  businesses. Join over 150,000 businesses that develop, manage and scale their applications with Digital Ocean. Head to do.co/changelog to get started with a $100 credit. Again, do.co/changelog.

• 1:56Speaker 02

  Welcome to Practical AI, a weekly podcast that makes artificial intelligence practical, productive and accessible to everyone. This episode breaks from our normal format just a bit. We are running Daniel's panel about AI for good in partnership with the R Data Science Conference that took place online this December. We hope you enjoy it.

• 2:22Daniel Whitenack?

  It is great to be back at our conference. So I spoke, I think, our conference in New York when it was in person. And I remember a heated discussion, since I was living near Chicago at the time, a heated discussion about various types of pizzas. And, you know, there were things said and all was in good fun, but it was a good time. So I missed, I missed that, but this has been a great experience. It's great to be here. We're going to discuss today how data professionals can engage with governments on AI for good projects. This panel will actually be recorded and released after the conference as well on the Practical AI Podcast. So if you're a podcast listener, there's a link there.

• 3:06Daniel Whitenack?

  You can follow along and listen again because it's going to be such a good time. Today, we have with us a couple of great panelists who have agreed to share their vast expertise with us. We have Danya Murali, a quantitative analyst at Arcadia Power. She's a data scientist with a passion for energy, environment and climate. And currently she's working on the data team at Arcadia Power, a fast-growing startup in the Washington DC area, seeking to create a 100% renewable energy future while saving customers money on their power bill. We also have with us Emily Martinez, who is an interoperability unit chief

• 3:50Daniel Whitenack?

  at the NYC Department of Health and Mental Hygiene. She has her master's degree in public health from Columbia's Mailman University of Public Health. And she currently spends her time connecting healthcare providers to the CIR in anticipation of the COVID-19 vaccine, which is of course very timely and also very much fitting within the AI for good space. So I'd like to maybe just start out this panel by asking, maybe we can start with Danya, just to get her sense of when we say AI for good or data for good or data practitioners working for good, what does that mean to you in terms of your day to day and things you've seen in your career?

• 4:34Speaker 00

  Yeah, thanks, Daniel, and thanks, Jared, for having me back. I am so excited to be able to be at this conference, even if it's Virtual this year. So yeah, AI for good. That's something that I care about immensely. What I think about when I hear just the term AI for good, it's the idea that we should be using data and we should be using AI for things that are equitable and helpful to communities kind of across the board. So not just one particular community or group of people, but kind of helping all people together and being very aware and cognizant of how you can use that data and how you can use AI to achieve that goal.

• 5:16Daniel Whitenack?

  And I guess a follow up on that. What do you think about the current state of how data and AI or data science is being used? Do you see it being used equitably now? Or are there places where that's not true?

• 5:30Speaker 00

  Yeah, that's a really good question. I think it's on a path of improvement, but still in a place where we should be all more aware of what impact our use of data and our use of AI can have. There's definitely, I've heard some scary cases of using things like facial recognition for like police related things, which could disproportionately affect communities' color and hurt various communities. So there's scary moments like that, but then there's also like really great things with like what we heard today with the panel with tracing COVID and using data and using AI to try and help just general health of our population. At Arcadia, we're more focused on trying to get energy, renewable

• 6:14Speaker 00

  energy as something that is just accessible to all people, no matter where your community's from. And also to do it in a way that we're making sure that we're not removing, but enhancing energy justice. And so that's something that we really care about. That's something that as a data scientist at Arcadia, we're constantly thinking about how our products affect different people and how we can help, how we can build for all.

• 6:40Daniel Whitenack?

  Awesome. That's exciting. I want to pose the same question to Emily to see your perspective on what triggers in your mind when you're thinking of using data for good or being an AI practitioner, working for good?

• 6:53Speaker 04

  Yeah, first, thanks Jared and the team for having me tonight. I think from my perspective, coming from a public health and local government bubble, I think we're always working for good. And we have in our fingertips access to incredible data, data that we can take action pretty quickly and impacts communities very quickly. So I think we're kind of on the receiving end too of outside private companies that work with government companies so that we're able to create and use newer technologies or methods. I know I view things a little different. I've, I've always been using data for good.

• 7:30Daniel Whitenack?

  Yeah, yeah, that's great. And you mentioned this sort of interplay between governments, whether they be federal or local governments and private entities. I was wondering, you know, of course, now that we're in this whole time of COVID and pandemic and we're getting to hopefully on the horizon, some distribution of vaccines, like, like was mentioned in your bio that, that you're specifically involved in that work. That necessarily involves, you know, a number of private companies. It involves logistics companies and all of those sorts of things. So I was wondering if from that perspective, you could give us a little bit of a sense

• 8:11Daniel Whitenack?

  of, at least from your own experience, how governments and private entities, commercial entities can work effectively together?

• 8:21Speaker 04

  Yeah, I think from, from my perspective and with the COVID examples, I think the collaboration has been great. It's been nice to be able to contract with outside companies. I know particularly we were kind of interviewing different vendors that we could partner with to establish a new point of dispensation system. So previously to working with these vendors, everything was on paper. So point of dispensing means we're set up to pop up, let's say, for example, a mobile clinic at a school where staff will be able to vaccinate people and people can just approach a school to get vaccinated. And all that needs to be tracked somehow. And previously this was all on paper.

• 9:04Speaker 04

  So that can still be happening in 2020. So it's nice that we're able, because of the, you know, severity and the impact that COVID has had, you know, worldwide, that we can move forward.

• 9:16Speaker 00

  With using better and efficient data products.

• 9:19Daniel Whitenack?

  Yeah, great. And of course, one of the underpinnings of this, like, like you mentioned, governments, whether they be national or local governments, have a lot of data that can be immediately put to use for certain AI for good purposes, whether that be as related to health, or I think as related to energy and other things. So, Danya, I want to kind of kick it back to you and hear from your perspective what the role of data from governments is in your own work in the energy industry.

• 9:53Speaker 00

  Yeah, we use a lot of government data. So, before I worked at Arcadia, I used to work at the Energy Information Administration, which is the statistical hub of the Department of Energy. So, it's been a, it's been a cool transition to go from like the government agency that collects the data to like being a private entity that uses the data. But one really big, one example I can give about how we use EIA data for good is we try and give our customers an understanding of what their CO2 impact is of their monthly residential electricity use. So, sort of like being able to keep track of your carbon footprint, but in a very, very precise way that is very much related to like your individual usage.

• 10:35Speaker 00

  And we do that using EIA data. So, the power of government is you have the ability to go and collect data and sort of like mandate that the data that your surveys get responded to, which is great. So, one example is EIA has a survey of all the combustion power plants across the US, so all the coal power plants, natural gas, oil, all of those power plants, and how much of each fossil fuel is being used in a year for each of those, and also how much CO2 is being emitted. And so, using that data set and using our individual customer data, we're able to create individualized forecasts and calculations of how much CO2 is released by using a certain amount of electricity,

• 11:19Speaker 00

  and then also how much CO2 is averted when you're able to source your power from places like wind and solar. So, there's a very direct relationship between how we use government data and our data to. Achieve that.

• 11:31Daniel Whitenack?

  Yeah, that's awesome. It's cool to see how granular you can be with with some of that thing, some of that stuff, and see how it affects individual people's energy usage and maybe their own mindset around that, which is really cool. We had a question from our audience, which I'd like to propose here. I think it's a good one. So we're on the topic of data sets and how some of these government data sets can be used with, you know, great success for AI for good projects. I was wondering, maybe we can start with Emily in the healthcare space. What are some sort of go to resources out there for people that are maybe wanting

• 12:12Daniel Whitenack?

  to either contribute in the healthcare space or look at data from the healthcare space? Of course, some of that is, you know, has some privacy concerns and all of that with that. So what's what's the situation there in terms of data that people can access and data that needs to be protected in certain ways?

• 12:30Speaker 04

  Yeah, that's a great question. I think right now there's a lot of data that's been public and easily accessible. I know in New York City, there's the Open Data Platform where different data sets from actually different agencies, including from the Health Department. So there's a range of things. I think they're pretty much up to date. There might be a lag in how recent the data is, but there's actually quite big variety of data that can be used by anyone who just wants to play around with the data. I think there's also New York State's public data sets and other states as well. So I think right now data is very much accessible to anyone who wants to take a look.

• 13:12Speaker 04

  In terms of security, yes, there's a lot of security, especially patient demographic information, all of that. Most of the data that's cleared is probably aggregate or has no way that can be connected to a particular patient. So that is very much tapped very tightly

• 13:29Daniel Whitenack?

  Yes.

• 13:30Speaker 04

  within the health department.

• 13:31Daniel Whitenack?

  Yeah. What about in the realm of energy, Danya? What's the situation in terms of, it sounds like you have this tool where people are able to utilize your analysis to kind of understand their own energy use. What about maybe for participants in this conference who are interested in maybe coming up with their own analyses or doing a little bit deeper study as related to energy? What is available out there for them to potentially utilize?

• 13:59Speaker 00

  Yeah, there are a lot of great government and otherwise data sets surrounding energy. Two things I highly suggest, so one, I sound like a broken record, I would highly suggest you go to eia.gov where you can see that they just collect a ton of data about all energy across the board, but also going to the EPA website, they have a really cool green energy calculator where you can put in like how many miles you drove or how many like airplane rides you took, which I'm sure is not very many this year, and what that, how that lends to like CO2 emissions and other environmental factors. Another place I would also suggest is just going to Kaggle. I don't know if you guys have used Kaggle before, but that is a really great place to get a lot of different types of data, both like healthcare data and energy data

• 14:44Speaker 00

  and pretty much any type of data you want. And also one of the things that's really nice about Kaggle is that oftentimes people have already done analysis and it's a very open source place where you can contribute to other people's things or pull from other people's things. So that's a, that's a really good source.

• 14:57Daniel Whitenack?

  Yeah, yeah, those are great. And I'll kind of throw in my own contribution here. So I work mostly in the language space for an NGO and there's a lot of great language and speech data out there. In particular, Mozilla has done an amazing job with Common Voice, which is a large data set of transcribed speech, which is out there in all sorts of languages. And then there's a whole bunch of open data that you can use for like machine translation projects. If you search for OPUS, which is an open parallel corpus, you can download a bunch of that. And yeah, I would encourage people to dig into that. It's pretty interesting when you start doing some natural language processing on languages

• 15:39Daniel Whitenack?

  other than English, especially because for one, it can help. Benefit, tools and support for languages. But also you run into all sorts of things fairly quickly with scripts other than Latin script and, you know, really long words or languages that don't use spaces. And so you have to think about all sorts of interesting problems. And so it's also really interesting from that perspective. While we're still on the topic of data and using data for good, I know that one element of this is also making sure that we're responsible, no matter what project we're working on,

• 16:17Daniel Whitenack?

  making sure that we are using data in a responsible way and not either showing bias against certain populations or possibly even unintentionally doing things that might harm people. So, Danya, do you have any thoughts on that front in terms of in your own work or just what you've seen in practice across industry, some good practices or things that people should keep in mind with respect to that?

• 16:46Speaker 00

  Yeah. So, one thing that I strongly advocate for is to be doing race-conscious data analysis instead of like racial-blind analysis. So, using things like zip codes can very, very easily be a proxy of race and socioeconomic status, but that doesn't necessarily mean that you should not look at those things. Like, just covering your eyes and being like, "I'm not gonna look at race, I'm not gonna touch race," is not the way to do it. To make sure that you are being equitable and you are really thinking critically about how your data analysis affects different people, you need to really be thinking about who those people are, where they come from, what their needs are. And to do that, you should be looking at things like race and like zip codes and the

• 17:31Speaker 00

  relationship between those things when you're doing your analysis.

• 17:35Daniel Whitenack?

  So, when is the time to do that? Is that when you're sort of doing your initial preprocessing and setting up your project? Is that like while you're doing your analysis? Is that afterwards and monitoring? You know, where can that sort of fit in?

• 17:49Speaker 00

  I think it fits in in many places. So, I would say that it should be kind of like in every step of the analysis. Like, when you're collecting data, making sure that you're collecting representative sample, for example. When you're doing the analysis, making sure that you're not making decisions or taking averages or doing things that are biased. So sort of like in every step of the data analysis lifecycle, being cognizant of that. I can give a better, a more particular example is at Arcadia. One, we recently started a chapter, a diversity, equity and inclusion chapter. And that contains people from all across the organization. So we have engineers, we have data scientists, we have people that work on the member experience team.

• 18:32Speaker 00

  And we all come together and we talk about how our different products relate to diversity and inclusion and how the data that we collect and the data that we use to create products affect that. And so by having and also like an outside, outside of your day job, like outside chapter that looks into these things and considers these things is also good. Like having checks in place to make sure that we are looking at different types of like data, diversity data. Another thing that we've been doing recently is just looking at the demographics of our customers to make sure that we're not over indexing in a certain population. And if we are, like applying the corrective features. So it's definitely one of those things where you can, you can mess up and that's okay and you can fix it.

• 19:17Speaker 00

  And if you're just aware of it through all the different steps of your data analysis process and your product development process, I think you could really do some good.

• 19:25Daniel Whitenack?

  Yeah, I appreciate that perspective. I kind of want to return over to Emily to the healthcare space. I know one of the things that you mentioned when, when we were talking prior to the panel is your work, you know, connecting patients to services during this pandemic. And this maybe connects to one of the questions we got as well from the audience, which is, you know, I guess in that scenario, there's also this element where certain populations, certain demographic factors have been shown to, you know, have higher risk and higher concentrations of COVID, higher death rates, all of that sort of thing. How do you balance like collecting and using sensitive, maybe racial information or demographic information when maybe some people are?

• 20:12Daniel Whitenack?

  You know, might not want to give that information, but you you might want to utilize it particularly in healthcare for certain purposes. What questions go through your mind and how do you handle some of those sensitive pieces of data, making sure that you're not, you know, exploiting those or making them, you know, gathering them when you shouldn't, but also gathering them when you should for for good reasons?

• 20:35Speaker 04

  Yeah, I think the city already has a good picture of where those areas are. So, reaching out to the community where with low resources, socioeconomic status, a lot of programs are developed to help those particular communities. So, I think we kind of already know where that is and a lot of the data reflects, we use a lot of data to find where these disparities are and they always align. Wherever there's a high index, it also matches poverty numbers. So, there there is a, you know, a good match correlation on that. So, I don't know, in terms of sensitivity, I think the data has, I don't know, we always

• 21:19Speaker 04

  use the data carefully and we already have these ties to the community. So, I don't think there there's a problem that we might backfire in our communication or, I think there's a lot of trust in that sense.

• 21:32Daniel Whitenack?

  Yeah, I think that's a really great point in terms of the connection to the community and not creating projects sort of without any communication and trust between you and the community and then sort of forcing a solution in and saying, hey, this is gonna, you know, fix all of your problems. There needs to be an open line of communication there and I think that's a really good point. So, I think we're pretty much out of time, but I do want to just give one more question to both of you. We can start with with Danya. I would just be curious to know, you know, what excites you about the future of using data for good? What excites you about the potential there and what impact it could have?

• 22:12Speaker 00

  Yeah, I love data. I also especially love art. One of the things that I'm really excited about that has grown in probably the last decade is the use of open source data and open contributing to software like art and then also conferences like this and different organizations like Our Ladies that really kind of try and get more people of various types of communities to come in and be data analysts and like make it so. It is accessible across the board, like things like General Assembly and these other types, or like Code Academy, these online classes where you don't need to get a degree to be able to do data analysis and have an impact and bring your own like personal bit to it.

• 22:55Speaker 00

  You can go off and do that online. And so it's, it's, I feel like a lot of the barriers to being data literate and being able to make really smart decisions and choices has lessened. And I think it's going to continue to continue to lessen. So it's, it's very exciting. I think it's an exciting time to be in this, this data space.

• 23:14Daniel Whitenack?

  Awesome. Yeah. What about you, Emily? What, what are you excited about?

• 23:17Speaker 04

  Yeah, I agree. And I, I echo everything that Danya said. I think also particularly in, in local government, there has been an increase in using open source tools, using R, particularly in the health department. I think it was a couple of years ago where we got an R server and we were all excited about that. And that has really pushed a lot of new R users. We've been mostly SaaS users. So I think we're going in a really good direction. And there's a really good big interest of within local government about data science. And it's just been very accessible. And I think that the biggest part is a sense of community in every sector. Data analysts have some forum, they can reach out and, and

• 23:58Speaker 04

  ask others like them if they have any questions. So I think that that's been the biggest thing, that sense of community.

• 24:06Daniel Whitenack?

  Awesome. Well, uh, thank you both for, uh, such a great discussion. Um, I really enjoyed it. Um, as a reminder to everyone, this will be published again on the Practical AI Podcast. Uh, check that out if you're into podcasting. Thank you again to the R Conference for making this happen. Um, I'm really glad that this conversation happened and this content will be accessible and, uh, enjoy the rest of your conference.

• 24:35Speaker 02

  Come hang out with Daniel, Chris, and hundreds of other AI practitioners in our community Slack. It's a cool place to be, not a lot of noise, some great signal, and best of all, it's totally free. Check it out at changelog.com/community. And don't forget to follow the show on Twitter for AI news and links, highlights from past episodes, and more. We are at Practical AI FM. We love to have you following along. Thanks to Daniel and Chris for hosting Practical AI, week in and week out, to the mysterious BrakeMaster Cylinder for the excellent beats you hear on all Changelog podcasts, to our sponsors who have our back, Fastly, Linode, and Launch Darkly, and to you for listening. We appreciate your time and attention. That's all for now. We'll talk to you again next week.