---
id: "workshop:the-basics-and-the-breakthrough-opportunities"
type: "InnovateUS workshop"
title: "The Basics and the Breakthrough Opportunities"
url: "https://innovate-us.org/the-basics-and-the-breakthrough-opportunities"
date: "2026-10-08"
duration: "60 minutes"
series: "AI in Public Health"
instructors: ["Lawrence Madoff","Andrea V King"]
moderators: ["Samuel Scarpino"]
recording: "https://www.youtube.com/watch?v=Z2H2NWmjbkg"
transcript: "automatic, from the recording"
source: "InnovateUS (innovate-us.org)"
licence: "CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org"
---

# The Basics and the Breakthrough Opportunities

InnovateUS workshop · 8 October 2026 · 60 minutes

Public health is one of the most important—and underdeveloped—frontiers for artificial intelligence. While AI is rapidly reshaping clinical care, its application in public health remains early, fragmented, and full of untapped potential. This session introduces participants to the current state of the field and provides a practical foundation for understanding how AI can support population health goals.

Participants will receive a clear, accessible introduction to core AI concepts and explore how they connect to established public health methods including epidemiology and real-world evidence. The session will examine approaches for developing ethical and responsible AI systems that center human and community agency and discuss how public health leaders can begin building AI literacy and readiness within their organizations.

The workshop will also address why adoption has moved slowly despite growing interest. Participants will explore common barriers—including fragmented data systems, weak cross-agency coordination, institutional risk aversion, and complex governance environments—and leave with a practical understanding of why advancing an AI-enabled public health agenda has become increasingly urgent for improving outcomes and reducing harms such as misinformation and inequitable access.

## Speakers

- **Lawrence Madoff**, Medical Director, Bureau of Infectious Disease and Laboratory Sciences, Massachusetts Department of Public Health
- **Andrea V King**, Washington State Department of Health, and Founder, PubHealthAI Collaborative Network

## Moderators

- **Samuel Scarpino**, Professor of Public Health and Health Sciences and the Executive Director of the Institute for Experiential AI at Northeastern University

Part of the series [AI in Public Health](https://innovate-us.org/workshop-series/ai-in-public-health).

Recording: https://www.youtube.com/watch?v=Z2H2NWmjbkg

## Transcript

Automatic transcript of the recording: names and terms can be misheard, and the times can be a few seconds off. Each time links to that moment of the video. The transcript alone: https://innovate-us.org/the-basics-and-the-breakthrough-opportunities.txt (plain text) and https://innovate-us.org/the-basics-and-the-breakthrough-opportunities.json (every segment with its start and end second).

**Samuel Scarpino**

[0:00](https://youtu.be/Z2H2NWmjbkg?t=0) I'm really excited about today's session. My name is Sam Scarpino. I'm a professor at Northeastern University in Boston. I've spent the last 15 years working in translational public health and global health security, both on the research side of how you integrate new technologies into public health, but also in the translation and implementation side. I'm joined today by two colleagues who have similar backgrounds in terms of the work they've done bringing new technology into the public health space. Larry Madoff, who's the medical director for the Massachusetts Department of Public Health, and Andrea King, who's at the Washington State Department of Health and leads and founded PubHealth Futures and the PubHealth AI Collaborative Network. Just before I turn it over to Larry and Andrea for introductions, I think by way of broad framing, and I probably don't need to convince this group that public health has been one of the most important drivers really in human history of individual health, community health, economic prosperity, happiness, and wellness. Throughout the history of public health, we've seen the regular integration and adoption of technology that has unlocked new levels of health and wellness that were previously unobtainable, whether we're talking about vaccines or sanitation and many things that have come since then. One of the challenges, of course, is that both public health and technology, they move at the speed of trust.

[1:35](https://youtu.be/Z2H2NWmjbkg?t=95) We have to be very thoughtful, very deliberate, and center ethics and responsible use in all of the technology we do, especially in the context of artificial intelligence. What we're going to be talking about today is both what's real, maybe what's less real, where we think things are going in the context of AI and public health, but just as much talking about the challenges and opportunities of how we bring this powerful technology to bear on improving the health and wellness of our communities and populations, but in a way that's responsible, that centers data sovereignty, ethical and responsible use, and really works to advance health at any address. With that, I'm going to turn it over to Dr. Mattock to introduce himself and say a little bit of an

**Lawrence Madoff**

[2:24](https://youtu.be/Z2H2NWmjbkg?t=144) introduction. Clay? Thank you, Sam, and thank you for having me today, and thanks to the organizers also. I'm trained as an infectious disease physician. Actually, my background is in medicine and to some extent in laboratory investigation. My early part of my career was as a laboratory investigator and academic infectious disease physician. I continue an academic role and a clinical role at the University of Massachusetts in Worcester, where I'm a professor in infectious disease attending. Then the other hat that I'm going to be talking mostly about today is my involvement with Beacon, a project based at the BU Center for Emerging Infectious Diseases that's led by Mahid Bhadelia. I'm going to be talking about AI actually in all of those spheres, but primarily to do with Beacon. I'll just mention that as a clinician, we are offered the opportunity more and more to interact with AI on clinical questions.

[3:46](https://youtu.be/Z2H2NWmjbkg?t=226) This is a space where accuracy and, I guess, curation of the artificial intelligence information is particularly appropriate. It's especially important, I'm sorry, because you want to know that the information that you're looking at is not just a Google search. You're actually seeing information from authenticated evidence. In that sphere, I have been using, in particular, open evidence and also UpToDate, which is a commercial product, a commercial subscription product that many clinicians have used for access to updated, curated health information. In the public health space, for many reasons, I'm fortunate to be in Massachusetts and a state that has really been a leader in public health since Paul Revere was our first public health commissioner in Massachusetts going back a few years.

[5:05](https://youtu.be/Z2H2NWmjbkg?t=305) One of the things that the Commonwealth has done recently is to roll out chat GPT accounts to every state employee, actually, everybody in the executive branch. That has been an exciting innovation. Along with that, it's being rolled out. It's not completely rolled out, but it's being rolled out across different speeds, across different departments and agencies within the state. Along with it comes pretty robust training. I was happy to see that part of our training comes from Innovate US. That was so-so the people who are bringing you this presentation today. Lastly, I'm going to talk about BEACON. BEACON is a system that mines the internet for information regarding new and emerging infectious disease outbreaks.

[6:15](https://youtu.be/Z2H2NWmjbkg?t=375) That's its primary role. It has some other functions as well, but primarily, that's what it's built on. It's based at BU Center for Emerging Infectious Diseases. I think what is particularly innovative about BEACON, and I'm proud that I had a role in helping it come together, although, as I said, it's being led by others at this point, was in bringing together human subject matter experts and an expert AI system to try to monitor for outbreaks of emerging infectious diseases and to develop, generate reports on these disease outbreak situations that would then be vetted, edited, verified by humans prior to their being made public on a website and via email. I don't know if someone can, you may be able to see in the chat the link to beaconbio.org, but it's free and open access.

[7:38](https://youtu.be/Z2H2NWmjbkg?t=458) Also, I think importantly, open source, the machinery that runs BEACON, what we call the pandemic LLM, which was built on the back of a commercial LLM llama and then was actually trained on a huge corpus of infectious disease information. This was done at the Hariri Institute for Computing at BU and led by Yanis Pascalidis, who was a public health informatics guru there, I guess, I would say a leader in the space. He and his team have put together what I think is a remarkable tool, this pandemic LLM, which, as I said, is the machinery driving BEACON and interacting with the human subject matter experts. The training was in two phases, actually initially ingesting human health information, some 6 billion tokens of information from a variety of sources that were selected for their reliability and authenticity.

[9:08](https://youtu.be/Z2H2NWmjbkg?t=548) Then, additionally trained on a corpus of animal health data because people who work in the field of emerging infectious diseases realize that zoonotic disease, veterinary disease is an incredibly important aspect of emerging infectious diseases. It was very important that the pandemic LLM also be trained on animal diseases. That was an additional training that went on. Then there has been a whole lot of other work that's gone into the construction of this system that's probably beyond the scope of what we're going to talk about today, but including the vetting of sources, being able to grade and extract information from media sources, from official government sources and other sources of information, rating them in terms of reliability of the source and urgency of the report to help the human subject matter experts decide on what to review and what to update. I am going to stop there and turn it back to Sam. I know there will be questions. I'm happy to be there to answer.

**Samuel Scarpino**

[10:33](https://youtu.be/Z2H2NWmjbkg?t=633) Great. Thanks so much, Larry. Just as a reminder to everyone, we'll have questions that we'll take from the chat. If you have questions, please put them in the Q&A. We'll get to those towards the end of the hour. I'm excited now to turn it over to Andrea King to provide her introduction, tell us about her background, and also about PubHealth AI

**Andrea V King**

[10:57](https://youtu.be/Z2H2NWmjbkg?t=657) and the work she's been doing there. Andrea? Hi, everyone. It's great to be here. Thank you to Innovate US and Dr. Scarpino for inviting me here to speak with you today. I'm Andrea King. I am a public sector tech person, I suppose. I'm trained as an epidemiologist and statistician, and I've had a career spanning academia and working in several large health departments here in the US, plus some global health work that's part of some international ongoing work that I'm doing. I spend a lot of time in New York City dealing with every outbreak that passed through the borders of JFK for close to a decade. More recently, I've been with the Washington State Department of Health working on data modernization initiatives. So that journey from working as a junior analyst in New York City up through working with their biggest disease surveillance systems and then moving to Washington led me on the course of further upstream from the data consumption spot, working in ways that would affect system change more than just cleaning the same data set over and over. How do you improve data systems?

[12:22](https://youtu.be/Z2H2NWmjbkg?t=742) When you start digging into that, you look further at how do you standardize data systems? How do you bring people together who need to fix these systems? Ultimately, that it's a tech problem and it's a people issue and it's a capacity issue. So I started the PubHealth AI Collaborative Network, which I'll get into here shortly, to bring people together around the adoption of new technologies in the public health space. So how do we do this for public health benefit? Leverage AI and the things that come after in a way that's really going to serve our field across a number of different facets. So I'm just going to quickly show a couple of slides so that those of you who are not yet familiar can get a sense of what I'm talking about here. So the collaborative started in June of 2024, and it was an outgrowth of some work I was doing with Google's public sector rapid innovation team on a sprint, fine tuning the Gemini model for public health use cases. And the intention was just to have a couple of small group meetings for a few months to share the findings from what we had been doing in the Pacific Northwest with other colleagues that were in state health departments that surrounded Washington and some academic partners.

[14:04](https://youtu.be/Z2H2NWmjbkg?t=844) But as I was sharing results, we were finding really interesting things that even three years ago, most of the models were credible at passing the certified public health exam out of the box with no additional fine tuning or training necessary. So we started applying them more broadly internally to use cases in a group concept. And people asked if they could forward the network and invite others. And pretty soon I realized I had something that I tapped into here a bit with this experimentation around applied AI and public health. So now we're running around 250 or so institutions globally with 700 or so members. It's really had a great year for growth and collaboration in a way that Far exceeds all expectations on my part. But I really, I like bringing people together to figure out new ways to solve old problems. The same things that have dragged down our efficiency in government for years are things that affect all of us in many different ways across the sector.

[15:20](https://youtu.be/Z2H2NWmjbkg?t=920) And I want to, you know, let people struggle together more than struggle alone, I think, for these things to learn from those who went first. People have had good use cases for applied AI and public health. I gather those routinely, highlight them, share them out, make sure that the work that's done in, you know, jurisdictions around the country gets amplified and shared and spread to build that knowledge and capacity nationally and increasingly internationally, because a lot of the challenges are shared, even if the forms are different. And I, we host bi-weekly meetings to bring in people from academia and industry, and to highlight use cases from members of the network who have something interesting to share on their findings, again, to foster that collaborative effort towards these things.

[16:26](https://youtu.be/Z2H2NWmjbkg?t=986) And we've written papers together, we've developed policy packs, we do vendor evaluation work. A lot of this is just a benefit of conversation from people. I connect people who are working on the same topics and may not even realize it. So there's a value to that. And then finding a common ground with others on trust and ethics makes a big difference, because this is a network that's curated around public sector participation. This is not an industry network, so it's really meant to be peers learning from each other, but in a cross-cutting way that brings in the best of each of our fields and aspects. And Pub Health Futures is a side project that I started to basically let us work a little harder for people in the network who are wanting to know more, just needing more specific support. It's a social enterprise, and it's centered around being able to work one-on-one with agencies in a way that kind of just adds capacity to what we do. And if anyone out there is curious or interested, we have a lot of free open access resources. You can join up to the biweekly calls if that's of interest. And we have a YouTube channel with about a year and a half worth of archive at the moment of the biweekly talks on everything from folks from DeepMind talking about bias and evaluation and AI models to demos that I've spun up just to show new tool capabilities. So a lot of varied content there that may be of interest to folks. And then if there's anything I can do to help you out specifically, please just reach out and let me know.

[18:25](https://youtu.be/Z2H2NWmjbkg?t=1105) Sharing that, and I'll pass it back to you, Dr. Stokina. Thank you.

**Samuel Scarpino**

[18:30](https://youtu.be/Z2H2NWmjbkg?t=1110) Great, thanks. Thanks so much, Andrea. So as Bonnie mentioned, I've got a few questions that I'm going to ask Larry and Andrea, and then we'll have time at the end for questions that we'll select from the Q&A from the audience. So if you have questions, please put them in the Q&A. And for those of you who've joined a little more recently, if you'd like to introduce yourself in the chat, please do go ahead and do that. One of the things from my perspective that is a little bit misleading about artificial intelligence is that we spend most of our time talking about the technology. We talk about generative AI and neural networks and lots of acronyms. It sounds very complicated. Certainly aspects of it are very complicated. And less time talking about what it actually takes to make these tools work in ways that really deliver value to the populations we're serving and trying to improve.

[19:33](https://youtu.be/Z2H2NWmjbkg?t=1173) Larry, I want to ask you a little bit about part of that equation of making it work. One of the things that I think is true about your career is that you've been a part of both building and working on really interdisciplinary teams with private sector technologists and academics and policymakers and public health practitioners, community groups. What have you learned about how to bring those kinds of groups together and have everybody marching towards the same goal? I mean, is there anything you can tell us about what that's been like? And maybe also recently, if anything has changed with the AI work that you've been doing at Beacon in that context of

**Lawrence Madoff**

[20:15](https://youtu.be/Z2H2NWmjbkg?t=1215) these interdisciplinary teams? Thanks, Sam. It's a good question. And yes, all aspects of my work involves interdisciplinary teams where people have vastly different expertise, experiences, knowledge, approaches, and sometimes even values. And that's always challenging. And I think this doesn't apply just to AI, but to really any endeavor in the public or private sector. So understanding, so I would start with being able to listen, right? I think that a key characteristic of anybody who's trying to lead is to be able to listen to the people that they're working with and to try to understand what their experience is, what their expectations, what their knowledge is, and trying then to understand that. And then the next challenging thing is what I've sometimes called herding cats. And this is always a challenge, right?

[21:46](https://youtu.be/Z2H2NWmjbkg?t=1306) As people have their own individual aims, I guess, in anything that they undertake. And you need to corral these different aims into a common goal and into common expectations. And also have some guardrails in terms of, for example, ethical standards, privacy standards. Privacy is really, in both medicine and public health, is a deeply held value. And how do you safeguard privacy in the realm of AI has been a really challenging problem, I think. And how do you ensure that you're not sharing protected personal information with many people? And so I think getting everybody to coalesce around the goals and the values is really the challenge.

**Samuel Scarpino**

[23:04](https://youtu.be/Z2H2NWmjbkg?t=1384) Thanks, Larry. I guess building off of that, Andrea, one of the things that has been true for a while in the context of public health, but seems to be accelerating with AI, is the involvement and collaboration with the technology companies that are building these tools and releasing them. You mentioned working on training models in this context in the past. There's one element of this, which is sort of how we teach people about what matters in public health from the actual use cases. But then there's the other component you touched on, the ethics and values piece, which is how we ensure that as we're building these tools, we center the values of the public health profession, but also the communities that we're serving. So could you maybe just tell us a little bit about sort of how you've approached centering the ethical values and the community norms in the technology work that you've been doing and in your interactions with technology companies? Yeah, I mean, it's a complex topic

**Andrea V King**

[24:16](https://youtu.be/Z2H2NWmjbkg?t=1456) because there are so many different angles to the perspective on AI technology at the moment. You get the push and pull in the public health space, particularly because on one hand, these technologies are immensely capable in a diverse series of tasks, a lot of different ways that we haven't really dealt with that in the past. Tools were more single function, or you had an analytic tool that you would write code in to do a number of different things, but it didn't translate your press releases and do your analytic work and also write your code and then keep visibility of the data pipelines all from the same type of platform. And so within the field of practitioners, there's a lot of different uses and needs and tug around that. And then if you look at the wider perspective of the community, people will have informed hesitation about the environmental impacts and the community level outcomes that may be adversely associated with AI data center and noise pollution and things of that nature. And so I think I tend to just approach it from a very open-minded listening perspective that everyone's opinion is informed by their understanding that I have never been them, so I don't have that type of understanding. And so what can I learn from the people who are raising certain topics in how do I grow to understand that?

[25:54](https://youtu.be/Z2H2NWmjbkg?t=1554) How do I help other people understand that? And how do we have those conversations in a way that's productive and aligned towards this kind of open-minded listening perspective? And aligned towards this as mutual benefit. And I tend to look for opportunities to work with the technology companies that align around that, that mutual benefit and improved health outcomes. If they're really truly having a collaborative stance and trying to engage for technology for good, I think that tends to resonate more with the public health audience than just getting some credits, getting some licenses, and then not having that dialogue with the providers as much. So that engagement takes a lot of forms. Sometimes we'll do very specific topic-focused sessions at PubHealth AI, or we'll start working groups to try to figure something out that's on the minds of a lot of different jurisdictions. Like, how do you effectively engage your IT and privacy departments? That's been a big thread of conversation for a long time now. And I don't know that anybody has the perfect answer, but a lot of us have some good ones. And as those conversations start to grow, you do kind of form around consensus to some extent on, at least within the network, we have sort of a viewpoint on a number of these things that seems to be workable now. But as technology changes and as the capacity of these tools grows, we have to adapt that. So I think staying very adaptable and truly adaptable and truly trying to understand the technology for its capabilities is essential in informing that communication. If you're behind on things, you're not going to

[27:54](https://youtu.be/Z2H2NWmjbkg?t=1674) be having the conversation with the best information. That's the approach I take when trying to navigate some of these more challenging topics. Larry, did you want to add anything?

**Lawrence Madoff**

[28:07](https://youtu.be/Z2H2NWmjbkg?t=1687) No, Andrew, I thought it was a good answer. I couldn't tell with you coming off mute whether

**Samuel Scarpino**

[28:12](https://youtu.be/Z2H2NWmjbkg?t=1692) you were just ready for the next question or whether you had something you wanted to add. I hope that maybe some of you or all of you are seeing a trend here, which is that we're talking less about the specifics of the technology and more about how you actually make technology work much more broadly. One of the things that we talk a lot about at the Institute for Experiential AI where I work at Northeastern University is that there's really very little new about AI. There are some new capabilities that large language models have when it comes to working with human language and text. But for the most part, all of the challenges and opportunities with technology adoption are the same. It's moving a little bit faster. That makes it more complicated. There are hardware requirements that are constantly shifting that make it a little bit more complicated. But for the most part, it's coming down to good practices around monitoring, evaluation, research, and learning, co-creation, building interdisciplinary teams.

[29:14](https://youtu.be/Z2H2NWmjbkg?t=1754) The other component, and this is where I want to turn now, is data. All of the kinds of artificial intelligence that we use today, for the most part, fall under the heading of machine learning. Machine learning involves data. There's no machine learning without data. One of the biggest drivers of successful public health operations, data. One of the biggest challenges with public health, data. I want to ask Larry, the thing that I know you the best for is helping found organizations and also lead organizations through digital transformations, through bringing in new kinds of data into the public health space. You were integral to a lot What was happening, you know, moment by moment during the pandemic inside of the inside of the department in Massachusetts with new tools coming online? What, what, what have you learned over your career about the data practices that you're finding yourself using a lot more in the AI era? What, what can you tell us about, you know, how you build good data practices, especially when things are changing so fast?

**Lawrence Madoff**

[30:22](https://youtu.be/Z2H2NWmjbkg?t=1822) Thanks, Sam. Yes, data is a, data is a challenge. And it's a, it's a challenge sometimes in that there's too little data and a challenge sometimes in that there's too much data to, to adequately, to be able to adequately synthesize and use in a meaningful way. In public health, you know, we come, you know, and I'm, I'm, I'm a little bit of a dinosaur, I admit, and I come from a, a time when, you know, much of our data was, was manual and written down and sent to us by, by postal mail or, or fax, and where we were, you know, counting things. So, you know, in an era where, where your, your primary data stream was a laboratory report or a, or a case report form submitted by, by, by someone.

[31:33](https://youtu.be/Z2H2NWmjbkg?t=1893) And we've come into an era and now where we are encountering many, many different kinds of data streams, both, you know, within the traditional public health sector, you know, electronic health records and electronic laboratory reporting and vital statistics and those sorts of things. But also a whole realm of new data streams like social media and, you know, Google, Google trends or, you know, mobile phone usage data or mobility data or, you know, you know, just vast quantities of different streams of data coming together. And we try to make sense of them and we do, we do it in different ways.

[32:34](https://youtu.be/Z2H2NWmjbkg?t=1954) You know, and at the Department of Public Health, we create dashboards where bunches of different data streams are assembled into a, you know, a user interface that people can, can click through and look at, you know, what's happening with viral respiratory disease this month, you know, or today or this week. And at the same time, where, where, where are the wastewater signals and what are the hospitalization trends and those sorts of things. And then in the realm of something like Beacon, we're, you know, mining the internet really, right. And we're, we're looking for, you know, media sources and individual, you know, reports from astute observers and, you know, government press releases and websites that are, you know, continuously updated and trying to make sense of those. And, and again, synthesize data in a way that will be meaningful and important.

[33:40](https://youtu.be/Z2H2NWmjbkg?t=2020) So I think, you know, what's, what's changed in terms of data is the vast numbers and types of data that we encounter and the challenges of synthesizing that. And I think that is, you know, that's, that has been an important role for this, you know, I wouldn't say new tool, but, you know, importantly improved tool of AI.

**Samuel Scarpino**

[34:10](https://youtu.be/Z2H2NWmjbkg?t=2050) Yeah. Well, my, my understanding is that respiratory signals are going up. All right. So it's probably time for people to get vaccinated.

**Lawrence Madoff**

[34:17](https://youtu.be/Z2H2NWmjbkg?t=2057) Get your flu shot, get your COVID shot. Yes. Public health message, my public service message today.

**Samuel Scarpino**

[34:24](https://youtu.be/Z2H2NWmjbkg?t=2064) One of, one of the things that I use AI for probably the most these days is, is data integration and data cleaning. That really is a new capability that, that these large language models have unlocked for us. So I mentioned a few minutes ago, there weren't a whole lot of net new things that AI was, was able to do. It can do it faster. It could do it cheaper. It could do it for more people, but really this, this data integration piece is really kind of something that almost feels like magic. Although that's why you got to make sure there are humans in the loop and maybe you have a lot of good practices in place because you don't want to treat it, treat it like magic. You want to treat it like a tool that you, that you use and understand. And so I think that data integration piece is, is one where there's a lot of opportunity. And I know you said that, you know, pen, paper, Excel, fax machines were in the dinosaur era, but certainly I still encounter those kinds of data, not every day, but definitely every month. And I'm sure plenty of colleagues on the call, you know, here have, you know, have experienced that. And that is a place where AI can help, help a lot.

[35:34](https://youtu.be/Z2H2NWmjbkg?t=2134) Andrea, one of the things that you've talked a lot about, and you talk a lot about in PubHealth AI is a place where there are really accelerating and compounding challenges in the context of data. And that's like, that's how we protect the sovereignty of the data and, and try to help ensure that the communities that, that are represented in the data are accruing benefits when those data are used in AI, especially for public health. And, you know, we'd love, love to hear what you're, what you're learning from the convenings around how people are thinking about data sovereignty and any advice you have on, you know, for individuals who are, who are starting to navigate some of those challenges, but also opportunities in their own, in their own work or jurisdictions.

**Andrea V King**

[36:20](https://youtu.be/Z2H2NWmjbkg?t=2180) Yeah, data sovereignty is a perennial question that comes up often in the context of tribal nations and being able to have autonomy over tribal members' data when it's so enmeshed in many federal reporting data sets and census records and, you know, tribal, tribal nations are sovereign nations and it's government to government relations with, you know, tribal nations. With state departments of health or federal offices and being able to capture data on, on their, their tribal members back from government is, is a challenge. So, you know, being able to enter into data sharing agreements or there's a lot of procedural work that has to happen for tribal nations to get their data back. And I feel like that's generally one area where we could, could do much better as government entities and being partners to tribal nations and helping them access their data and understand it.

[37:29](https://youtu.be/Z2H2NWmjbkg?t=2249) But there's, there's a bit of a, an interesting groundswell happening among, you know, tribal nations that are, some of them are very forward thinking and moving in the technology space. And, you know, there's, there are a lot of developers who are working, like who are, who are tribal members themselves working to develop tools that, that speak to those needs within their communities. And, you know, data platforms that are AI driven that help identify and standardize the way tribal members are represented across, like, you know, the, the ACS, like five-year census estimates and like all these federal surveys that you can, with the right approach, aggregate and standardize that so they can give a more comprehensive rounded data set back to the, the nations that, that they, you know, that they're, they're working with. And that's, I think that there's, there's a lot of power to that as somebody who truly understands the challenges being part of that development process and that technical, you know, bring about the technical perspective and the, the, the perspective of the communities. And I think I just, I like co-development as a, as a mindset for a lot of these challenges, because I think that's the, probably the best way to solve them. If you come in externally with the assumption that your tool is going to fit, if you just tweak a few things, turn a few knobs, it's unlikely to really have that much impact.

[39:06](https://youtu.be/Z2H2NWmjbkg?t=2346) It's not a long-term standing that's going to give back to the people who are actually trying to use it. So, you know, that, that's one aspect. And I think, you know, I encourage people who are working with tribal nations in their jurisdictions to really have those conversations and try to figure out if there's a, a co-development path that can, you know, alleviate some of that, that procedural buildup and, and bring people closer to the data that they actually need to make health decisions for their, their constituents on both sides of that equation. And in the global health work I've been in, I think country level sovereignty is, is constantly a challenge as Dr. Madoff mentioned that, you know, we know diseases move across borders with, you know, incredible speed. And so does the information that's related to them. And that often the data transfer can't keep up, you know, from ministries of health, even within a country, the ministry of agriculture doesn't necessarily link up to the ministry of health. And, you know, while that's a, it's not a country sovereign issue, but those are, those are silos within a country that prevent the, you know, fundamental aspects of, you know, you've got a big gap in your understanding.

[40:26](https://youtu.be/Z2H2NWmjbkg?t=2426) If you're only looking at one, one piece of that, that puzzle. So I think that, you know, there's a, a case to be made for better integration of, of these data resources. And, you know, that's a, that's a One Health perspective. You know, if you look at animal and environment and human data together cohesively, you get a more rich picture of what might be going on. And so if you, if you kind of do that at a national level, then, you know, your sovereign approach to how you relate to other countries for data sharing and, you know, that kind of exchange is a bit more holistic. If you have everything kind of aggregated, you're not making deals with each ministry, with each, you know, research institute. And everybody's constantly trying to pull together aggregates from the same, for the same sources and reasons. So I think that, you know, with cloud technology being what it is now, you have the ability to store your data in any number of locations globally that bring it a lot closer to the source.

[41:36](https://youtu.be/Z2H2NWmjbkg?t=2496) And while it may not solve your immediate, like in-country sovereignty challenges, you know, you could at least house your data in Africa rather than in Europe. And, you know, from an ownership perspective, that, that matters. And, you know, just awesome data transit to it, that, that helps. But I think mostly that's a, that's a policy challenge, but there are technological ways that you could much more easily facilitate that, that kind of integration approach that would let countries share data more seamlessly within and across borders.

**Samuel Scarpino**

[42:18](https://youtu.be/Z2H2NWmjbkg?t=2538) Yeah, I think that, you know, one of the sort of grand challenges in the public health and global health space over the next few years is going to be how we, how we advance data and benefits sharing without looking towards retrenchment of, you know, siloing data. Because there's going to be a lot of, a lot of cases where data are used to disempower communities where data are going to be used and benefits don't occur. And that's going to incentivize people to stop sharing data. We're going to see a breakdown in collaboration. And so that means that we really have to have a coordinated effort to push for ensuring that we have, you know, communities of practice and standards and methods to ensure and protect sovereignty and ethical and responsible frameworks that, that carry our values forward. We're going to move over to questions from, from the audience. We're getting a lot of really great questions here. I had one more for Larry and Andrea, but I'm not going to ask it. I'm just going to answer it from my perspective, which is like, what are you really excited about from an AI perspective? Because there's a lot of doom and gloom and some, some of that's real, some of it's not real, but there are some exciting things.

[43:27](https://youtu.be/Z2H2NWmjbkg?t=2607) And one of the first questions is about something that I see a huge amount of opportunity in, which is, and it's targeted to Andrea, but Larry and Andrea probably both have thoughts on this. So what about AI for helping community members or, or healthcare professionals or others find benefits, get access to benefits, you know, engage in better care coordination? That to me seems like a really big opportunity space for AI and a place where I know there's already work happening. So, you know, Andrea, if you want to start to say a little bit about what you're seeing, and then Larry, I'm sure you've got some perspectives that you're seeing from the, from the state or from your clinical perspective, which sometimes is the same, I know. Andrea.

**Andrea V King**

[44:10](https://youtu.be/Z2H2NWmjbkg?t=2650) Yeah. I, I think that AI just in its ability to work across multiple languages at speed is really powerful in being able to provide that, you know, support for individuals who are interfacing with any type of entity government or clinical care in language of their preference. You know, it may not be perfect, but I think that, you know, people don't expect perfection, you know, from an AI translation at this point, but the idea is that is it, is it good enough, fast enough that you can give people information to make decisions that affect their health much faster than say, it takes three days to translate a press release otherwise. If you can communicate faster and more effectively, I think that's usually worth the occasional misfire on a colloquial translation or something. Generally, the translation services are very sound at this point. As far as helping people navigate benefits, there's a lot of work being done in helping people navigate, say, the federal insurances, Medicaid and Medicare benefits, and just figuring out what people are truly eligible for in these very Byzantine systems with lots of rules, AIs capable of ingesting just mountains of PDFs and standard operating procedures and benefits guidelines from any, you know, across an entire state or country, and then matching that up with, you know, what an individual may need or be eligible for. And those are things that can be done, like, on the back end. But I think that, you know, as a user, somebody who just goes to Gemini or Bot or whatever and types in a question,

[46:12](https://youtu.be/Z2H2NWmjbkg?t=2772) right, a lot of people just have that interface. That tells you where to start. And if they, if your AI tool, you know, it has access to your state regulations, it has access to all the policies from your health insurance company that are public, you can use it to dive into information and seek a bit more guidance on the thing that you're after rather than just sort of hoping that a Google search returns the top 10 search results that are going to be most effective for you. That said, you have to verify it, right? You know, you can't just jump with both feet in and say, okay, this is what the AI says, I'm going to go do that. So there's a, you know, there's a responsibility for your own well-being that always comes into play when you're using these tools to assess, you know, what you may or may not be eligible for. But there's a lot of upside in just chewing through vast amounts of information and being able to serve that out to people in their language of preference. It's a low lift, but I think it's a real game changer for people. Absolutely. Larry, what are you seeing? I have a concrete example,

**Lawrence Madoff**

[47:28](https://youtu.be/Z2H2NWmjbkg?t=2848) actually, where it was just a talking with some of our epidemiologists earlier at the state earlier today. And they have been noticing a steep increase in calls to our 24-7 epidemiology line. And they think it's attributable, and I believe them, to AI guiding people to where to go with a question. So, you know, I had a bat fly through my bedroom. What do I do? And AI, you know, doesn't leave, you know, if you Google search that in the past, you would get a list of, you know, resources. But now, you know, the AI response is, if you think you've been exposed to a bat, call your state health department, and this is the number. And so I think that it clearly is a tool that people are using and increasing their access to services.

**Samuel Scarpino**

[48:26](https://youtu.be/Z2H2NWmjbkg?t=2906) Yeah, I grew up in Indiana, where the bats are endangered. And so if there was a bat in your house, it was very complicated, because you had to call many different state agencies to figure out what you were supposed to do, including the health department. One of the questions that was in the chat, so we both, you know, Andrea mentioned sort of chewing through vast amounts of PDFs, and Larry talked, in your specific example, what the AI is doing there is chewing through lots of heterogeneous information on the internet to find out that what the state wants you to do is call this number if you think you might have been exposed to, you know, to a bat or to an animal that was acting abnormally. And, you know, so that we talked about data fragmentation and AI's opportunity there. Do you have other specific examples of, of how AI is being used to clean up and merge data? I don't know, Larry, maybe from say a little bit about what Beacon's doing, or Andrea, if you're hearing anything from your coalition of folks that you work with?

**Lawrence Madoff**

[49:26](https://youtu.be/Z2H2NWmjbkg?t=2966) You know, Beacon is trying to ingest huge amounts of data. I can't, you know, one of the dashboards on the, on the Beacon platform is just ingested news reports, and there are, you know, in the course of a day, there might be 500 ingested news reports that I could not possibly look at individually and categorize, and the AI is doing a first pass. What is this about? What's the source? What's the country? What, you know, and so it's, it's, it's, it's doing that job of ingesting a vast amount of information that a person couldn't, couldn't

**Andrea V King**

[50:20](https://youtu.be/Z2H2NWmjbkg?t=3020) handle. Yeah, I think where the, what I'm always thinking of is the book Rewiring Democracy, like there are four S's, speed, scope, scale, and sophistication is the kind of a framework for thinking about AI tasks, and that that's a, that's a scale question that is more, more data than a human can deal with, but you need to understand it, right? So I think that's a, that's a great use case for AI. One of the use cases I worked on with Washington was the team that was manually extracting records from a coroner and toxicology reports to, to capture circumstances of death and overdose scenarios. And they would build these narratives that were de-identified. So it was actually a great opportunity to work with an LLM against those narratives in a way that like gave public health value, but still protected privacy of the, you know, the people involved in the, in the records. And they would spend three or 4,000 hours a year typing and creating and coding these, these narratives into reporting for federal authorities for the, for the grants they were on. And, and it just, it takes time away from things where they were experts in something else and were really just doing a lot of data manufacturing to, to, to serve the needs of the program. So, you know, we found pretty quickly that LLMs can, can do credible work taking unstructured data to structured data. So I think that, that processing of, of text is, is, is a solid feature of a lot of these AI tools. And it has, as we were speaking about earlier, none of this is new. You could do this with regular expressions and, and, you know, fuzzy matching and natural language processing

[52:24](https://youtu.be/Z2H2NWmjbkg?t=3144) for years now, but the scale and the speed just makes it so much easier with LLMs to do that work. So I think that's a, that's an opportunity for applied AI in the public health space, anywhere where you have the ability to, to use narrative records that you need to get data out of for analytics, just across news or, or case reports or, I mean, electronic case reporting, I think is a huge opportunity space. And there's not a, AI is just starting to be integrated into that area, but I think there's a huge opportunity to, to make that more foundational to the way we do some of our data reporting.

**Samuel Scarpino**

[53:12](https://youtu.be/Z2H2NWmjbkg?t=3192) Yeah. It's one of the things that we've done a lot of work with at the Institute for Experiential AI at Northeastern, we did a project with the genomic tumor board out of the Jackson lab works with a rural oncologist where you have a new case that comes in and there's a test reports and emails and handwritten clinical notes. And all of that has to be integrated into an analytical database that populates PowerPoint presentations and other key resources. And this should sound a lot like epidemiological work or public health, you know, sort of day-to-day data management. And one of the things that's true about these large language models is that if the human sits down and provides a few examples, here's a PDF. Here's the numbers that I want to extract from the PDF. The LLM will be able to do that again and again, and again, even if the formats are slightly different and, you know, maybe evolves over time, as we all know, these, these data formats tend to evolve. So we're just getting up to the, to the end here. I want to thank Larry and Andrea for, for joining us. We have a final question for them on the way out, but I also want to thank our audience and the organizers from Innovate US and Bonnie and the team for, for all the amazing work they've done setting this up. And I want to encourage everybody, we have two more webinars in this AI and public health series. So on October 15th, we're going to talking about combating misinformation in the age of AI. We have two amazing panelists who are going to join us. One is both a scholar and practitioner of communication, but also runs a serious social media operation on the positive information side. And the other is a scholar of journalism and information. Then on the 22nd, we're going to have a conversation about benefits access and delivery with two individuals, including the commissioner of New York City health department, really from that sort of executive leadership level, how do you make these things work and how do you build the kinds of successful organizations and go through the,

[55:14](https://youtu.be/Z2H2NWmjbkg?t=3314) through the change management process. So I encourage you to join us for those, for those two webinars. Before I turn it over to Bonnie to close with, with a couple of housekeeping things, Larry and Andrea, number one, take home point for, for each of you, you want, you want folks to remember one thing. What is it? All right, I'll go first.

**Lawrence Madoff**

[55:41](https://youtu.be/Z2H2NWmjbkg?t=3341) AI is a, you know, an incredibly powerful tool and increasing in its power and capacity and scope and scale and all of those things. But it's a tool and it needs a human hand. It needs human oversight and guidance. And you know, without that we're, we're, we're not doing a service.

**Andrea V King**

[56:11](https://youtu.be/Z2H2NWmjbkg?t=3371) Andrea. Yeah, I would say don't wait, leverage the policies your institution has in place and get started because these things take time to socialize. There's a lot of change management that has to be done in organizations to get a, you know, the buy-in you need for long-term success. And the longer you take to try, just jump in and see how it goes because there's, there's a lot of upside.

**Samuel Scarpino**

[56:43](https://youtu.be/Z2H2NWmjbkg?t=3403) What I'll leave you with and then turn it over to Bonnie is, you know, folks working in public health in, in the broader allied health sciences, policymakers, community advocates. One of the things that all of us know how to do and we live and breathe every day is how to build trust and how to build coalitions that are going after improved health and wellness for the populations where we live and the populations that we serve. That makes all of you AI experts, probably even more so than if you were just pure technologists. So really lean into that expertise. Remember that trust is very hard to build and very easy to lose. And that AI is only going to be as successful as the trust that you build with, with the community. So again, I want to thank Larry and Andrea immensely, and I will turn it over to Bonnie for some closing housekeeping.

**Bonnie McGilpin**

[57:34](https://youtu.be/Z2H2NWmjbkg?t=3454) Thank you so much, Sam and Andrea and Larry for that great discussion and for everyone who joined us today. Please take the post-workshop survey. It'll pop up on your screen when the workshop ends. It helps us inform future workshops. And as Sam mentioned, we have some great workshops coming up. And next week we have three new workshops. We're starting two sessions, the good, the bad, and the ugly of predictive AI on Tuesday the 13th at noon. And then we have the first session of our using AI and public sector legal practice Wednesday at 2 p.m. Christine is putting more info in the chat. We hope to see you soon. Thanks again, everyone. And we'll close the

**Samuel Scarpino**

[58:15](https://youtu.be/Z2H2NWmjbkg?t=3495) session now. All right. We can all do the awkward wave.

* * *

Original page: https://innovate-us.org/the-basics-and-the-breakthrough-opportunities · Licence: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org
