John: [00:00:00] Welcome to The Flux, where we hear stories from people who have asked what if questions to better understand the world and talk about how data can help tell stories that impact decisions and create an intentional impact on the future. This is your host, John Courtier, CEO at Epistemics. In a world where the flux capacitor from Back to the Future does not yet exist, people have to make difficult decisions without always knowing how the future will play out.

Our guests are people who've taken risks, made decisions when uncertainty was high, and who have assisted decision makers. Transcripts provided by Transcription Outsourcing, LLC.

Welcome on to this episode of the flux. we're on with Sherwin Brown, who currently is at MITRE.

so Sherwin, glad to have you on the podcast. 

Sherwin: Thank you for having me. 

John: for most of our listeners, they might come from a really technical background or less of a technical background. [00:01:00] so in your case, what are some of the things that got you excited about agent based modeling in the first place?

Sherwin: I would say that,it's kind of a journey for me. andI got to MITRE in 2010 and my background was in health management and policy. Um,came in around the time when the Affordable Care Act was signed and, MITRE was given the contract to be the systems integrator. And, you get a lot of exposure in the policy side of the house, supporting CMS, the Centers for Medicare and Medicaid, 

and 

Sherwin: then, different projects, which are, very classically system engineering oriented, and working with stakeholders, bringing stakeholders together to solve, these complex problems.

and I always had like this interest in analytics and like data science and just empirical approaches, and that's just what I pursued and got myself involved in these different caggles, you know, might or host these caggles and people come in and we would like compete with one another, who's data science algorithm was like more predictive and stuff like that.

Okay. 

Sherwin: And [00:02:00] then, still keeping with the Affordable Care Act over the years. And, we recently just completed this audit and,and that's when I started getting my hands wet with like natural language processing, um, starting to get, some of the like data matching, semantic similarity, cosine similarity, that type of stuff.

And 

Sherwin: I was like, I think I really like this. I think I, and I can see how I can use this in a different way. and then came the 2023, um, CMS quality conference, which pretty much was a, was where I started seeing like, there's a problem and, we need the right tool to solve this problem.

And during this conference, the deputy administrator of CMCS was saying he was concerned about beneficiaries housing status, the Medicaid population, because during the pandemic, they had stopped disenrolling everyone, to keep people covered so that, they could. Have access, you know to vaccinations and primary care services and all that stuff Uh 2023 that's when they brought back disenrollments.

So all of a sudden you have millions of people suddenly disenrolled And the housing status [00:03:00] thing is something that in academia, they had talked about as well So I started asking questions about this inquiring, and like quite being more vocal about it is when I met,matt kohler and he introduced me to agent based modelingand so from there was the last year's conference, which is where we met.

and since then I've just been learning more and more about how these techniques can be used to solve complex problems. And the housing problem is a complex problem. and yeah, I think, everyone has a journey, but that's, that just happens to be mine. 

John: Sure. What were some of the, competition of like, The CAG was like more predictive than others, like that were fun that you remember from that time.

Sherwin: we did some, uh, risk scoring stuff and like the home health side. and so just looking at, trends in the data to see, is there likely going to be fraud, in this,in this, sector? And, what type of, measures do you look at? Which were going to be the most highly.

predictive variables [00:04:00] that would predict, there could be fraud within the home health context. That was one of them that we had done. 

John: What are some of the reasons why you think agent based modeling is a good technique for people to use to solve these policy and kind of complex challenges?

Sherwin: I think that anyone can look at a good Asian based model and understand the story that it's telling. And I think that's the difference maker. is that it's visual. It's almost like playing video games in a way. One of my favorite games as a kid was Sim City. And, I just liked being, having the ability to like change assumptions about where, these buildings would be located and, how like the commute patterns might be impacted.

And, stuff like that. But if you can see something that you like live in, I'm from Baltimore. I live in Baltimore. I drive around these neighborhoods. If you can see,a change in like homelessness rates, for example, by changing a few different assumptions and in the makeup of this ecological footprint, and see how that impacts,the outcomes of what you're looking at.

I think that people resonate with that, and [00:05:00] just we always, talk about here is like different tools, provide you a different insight. But for me, I think agent based modeling provides deeper insight and it's easier to understand. 

John: Yeah, the storytelling aspect's been really important for us and people we've worked with in the past.

bringing up the aspect of like being able to see your city, your town, your like neighborhood. we've actually had some policy makers say that's what enabled them to get policies passed because when they took a model of adoption of vaccines or impact of different programs they could roll out for the opioid epidemic and seeing it in Baltimore or seeing it in D.

C. or seeing it in Sacramento, California, they're able to rally people around it because, to your point, you can tell a better story for people. so,getting into agent based modeling recently, what are some of The key stories that you think are yet to be told where agent based modeling can have a good impact.

Sherwin: So I'm looking specifically at the complex relationship between homelessness and opioid [00:06:00] overdoses. The overdose rate in, nationally, is concerning. Between 2019 and 2021 there was a 97 percent increase in overdoses across the country. And looking at the city where I'm from, we're at 10 times that rate.

You know what I'm saying? and there is a complex relationship with housing status. and I think that story is not well understood. it's complex. But, it's not to say that it can't be represented, in a way that can help policy makers make decisions. And another thing about agent based modeling, And this complex problem is that this is something that can only be solved by coordinating across levels of government.

This is not something that is one person can solve it. you can't just point to one level of government and say, this is all your fault. It's not.

John: Yeah, that's a critical thing when piecing these models together, being able to represent here's the people that are impacted by this issue. You hear the decision makers, but here's where the resources are coming in from and figuring out that balance is It's going to be [00:07:00] key for any city or any state to figure out the opioid crisis in their own area.

 So, we're here at the Complex Social Systems Society Conference year two. Are there anything, or any, topics or people that you're excited to see again, or learn about? 

Sherwin: there's actually everyone that I saw last year and here I'm super excited about.

last year when I was here, I was very impressed by all the presentations. I think everyone here is doing meaningful work.a couple of people that, I'm looking forward to hear present, again this year, would have to be, Steve Kimbrough from University of Pennsylvania. last year he gave a presentation, and it was really about.

like the impact of like misinformation and that was, and like, you know how, when there's, when you are in a noisy environment, we tend to mistake each other. and so he had a net logo model, which was basically demonstrating. That is basically, you know, we're kind of like mistaking each other right now because we have, we're in very noisy environment.

Okay. 

Sherwin: yeah, this is a, this is another really hot [00:08:00] topic, misinformation, disinformation, and I'm looking forward to see what he has to say this year. 

John: Cool. Cool. So some people who tune into our podcast, they might be considering investing in agent based modeling, bringing it into their companies.

is there anything that excites you about where the field is going?

Sherwin: yeah. with companies like Epistemics, with what's going on with Mesa,I think we're building a community right now and we're raising awareness about these tools and how they can be used.

And also with like artificial societal representation, these are things that are being promoted by various philanthropies, as you're probably aware. One of them is the Bloomberg Philanthropy, and they have an annual conference. There's like an AI summit. They bring together different mayors like around the world that are testing different interventions out.

And, that's something that I'm excited about, I'd to see, I'd like to see how different, jurisdictions can, learn, from each other in tackling problems. And I think there, there will be a [00:09:00] need for more coordination between, like horizontally between like governments, but also vertically.

And I think these artificial societal tools can facilitate those conversations. 

John: So final question. This will be a bit of advice to some students that tune into this podcast. If you were talking to somebody in high school or early on in college, how would you,how would you ask them to think about getting into agent based modeling or what do you think a good path of getting to just agent based modeling might be for them?

Sherwin: I'm thinking about when I was a student, right?I had no clue I would be into agent based modeling at this time. I just generally think that as a We're lifelong learners, no matter who we are, whether we're in college or not. and the education, uh, ecosphere is changing, very rapidly as well.

I would say always follow your curiosity. and whether it takes you to agent based modeling or not,there's no problem. there's, agent based modeling is here. and there's plenty of, there are many things in life that you can be interested in. [00:10:00] But if you do find yourself interested in a new base modeling, I would say, find a good community to, have conversations with and learn from one another because that's the best way to learn is through these types of conversations 

John: for sure.

Yeah, our number one company valued epistemic is curiosity. yeah, definitely makes sense. Sherwin, thanks for being on the podcast as well. Really? Well, thank you. 

Sherwin: All right. Thank you.