← Priorities Podcast
Priorities Podcast
The AIDE initiative: How AI can help emergency managers better prepare for disasters
00:24:33
Greenberg joins the podcast to discuss the AIDE i…
Loading summary
Transcript21 lines
- [00:03]
A
Hello and welcome to State Scoop's Priorities podcast. I'm Sophia Foxwell, a reporter for statescoop. This week I interview Jeremy Greenberg, senior advisor for Aspen Digital, about the Artificial Intelligence for Disasters and Emergencies initiative, or aid. We discuss how the initiative is exploring practical ways AI could help state and local agencies better prepare for disasters without sacrificing public trust or accountability. But first, here are the biggest state IT stories of the week. The Federal Trade Commission finalized an order Friday against K12 software vendor Illuminate Education, directing the company to improve its data security measures and barring IT from misrepresenting its data privacy practices or breach notification times after A breach in 2021 impacted the data of more than 10 million current and former students. Minnesota is launching a sweeping effort to modernize the systems that support Medicaid, SNAP and other public assistance programs after lawmakers approved a $90 million package to support replacing the state's aging IT infrastructure and improving coordination between state and local agencies. The multimillion dollar investment comes as states across the US Race to meet new federal requirements affecting public assistance programs. This month, Kader Zakaria joined the Illinois Department of Innovation and Technology as the state's first chief artificial intelligence officer. A long time technologist, Zakaria will lead the department's enterprise AI strategy and support the state's responsible adoption of artificial intelligence. As communities across the United States brace for another active hurricane and wildfire season, a coalition of researchers, technologists and emergency managers is examining how artificial intelligence could help support agencies before, during and after disasters. Rather than focusing exclusively on generative AI tools such as ChatGPT, the AI for Disaster Emergencies initiative or AID, is evaluating a broad range of technologies. Tools like AI powered wildfire detection systems that can identify fire starts using infrared imagery and and flood monitoring technologies that can analyze, stream, gauge data and forecast where flood waters may travel. The initiative found that other AI tools can help automate administrative tasks, review emergency plans, develop training exercises and improve situational awareness during disasters. Jeremy Greenberg, senior advisor at Aspen Digital, part of the Aspen Institute, said that while AI technologies can be helpful tools, emergency managers must remain in control of critical decisions prevention, particularly during high stakes disasters such as hurricanes, wildfires and floods.
- [02:56]
B
So AID stands for Artificial Intelligence and Disasters and Emergencies and funded and founded by the Markel Foundation. About two years ago, the Markle foundation started doing some research just context. They have about a century of experience working in a variety of different public spaces, all the way back to helping the formation of Sesame street And then after 9 11, working on some national security Matters some upscaling of American workforces, and then then wanted to start a new initiative that would really lend experience and credibility in the emergency management space. So about two years ago, they started doing research, independent research with emergency managers, technologists, academics, to sort of figure out what this would look like. Obviously, we have seen this rapid explosion of AI capability. We know that natural disasters are increasing in frequency and intensity. So these two things coming together really generated what is now being called the AID initiative. So Aspen Digital, which is a program at the Aspen Institute and RAND partnered under Markle starting last October, formally and launched the initiative, which to date, and we can talk about this as we get into the podcast, but really have done an enormous amount of outreach to the emergency management community, state, local, private, voluntary organizations, and on the academic side, doing a variety of research behind the theories just to make sure that the science matched the antidotes, and then figuring out, okay, now that we know the problem, what can we do to help?
- [04:28]
A
Yeah, so what can you guys do to help? From my understanding, the initiative focuses less on developing new AI tools or software and kind of using the existing resources that are available for state and local entities. Do I have that right?
- [04:43]
B
So a little bit of both. So our partners at RAN actually researched all the tools that were out in the AI suites. I think this is a key point for any discussion around AI. It's not just large language models, right? That's a default. We think about, you know, chats and cloths and those tools in Gemini's. Those tools are amazing, but they're, they're just one tool in the suite of, of artificial intelligence capability. Think about predictive analytics, think about drug capabilities, think about computer vision, right? So really understanding what tools are out there and then how they might be adopted appropriately across use cases through preparedness, mitigation, response and recovery. So while that research was going on, our team over at Aspen Digital led a series of workshops, about 30 different workshops with a hundred different, close to 100 different jurisdictions at the state, local, private and voluntary level, engaging in basic questions about where are you seeing friction points just in emergency management? And then where are you seeing adoption of technology that might be able to help some of those areas? So combining our social research, RAND academic research into some ideas about how we might be able to move the needle forward and advance AI adoption in emergency management to meet some of the demands that emergency managers are having day in and day out.
- [06:01]
A
Before we get into some of the solutions that you've been able to come up with, what were those friction points? That you were hearing from these emergency managers, what were the biggest challenges that they needed help addressing.
- [06:13]
B
So I would say there are a couple and I spent 23 years as an emergency manager on the federal side, spent equally as long as a volunteer firefighter. So when you approach these conversations, you have some preconceived notions through your own experience about what it is that the friction point. So really trying to balance what we knew just from our own experiences on the team and hearing what people are having challenges with today. So I'll break it down a couple of key themes. So one is that there's always an exceeding demand beyond your time, right? Everything from individual community preparedness, getting out in the community, face to face conversations with your constituents about readiness, home safety, fire extinguishers, smoke detectors, go kits, communications plans, those kinds of things. That's where emergency managers want to be. But you have this demand signal for plan review, exercising grant development. And most emergency management offices in around the United States are much smaller than you would think. Most community level, local level offices are one to two people. And even larger cities and bigger states have larger offices but commiserate with their populations. So the demand of time, capability and then what the expectation is when a response happens, it's borderline overwhelms emergency managers. So part one is the overall demand against what your requirements are. The second, the challenge we saw is primarily around procurement. Buying anything in the public sector, whether you're local, state or Fed, is challenging. And a lot of the newer technologies, specifically around AI, large language models, are subscription based. Right. We all you use any of these tools, you go on, you can put your credit card in and you can buy it. Government procurement systems are not the same way. So how do you have that same rigor that we have from sedest response where we want to move, move, move quick by quick. But you're in your normal procurement cycle and parsing cycle. That's a challenge. And then the last is, and this resonates, I'm sure with everybody is sort of the cultural concerns around understanding how the system works, understanding where the human is in the loop, making a decision, and how we are leveraging technology to make ourselves more efficient and more capable. But acknowledging that the decisions are still being made by human judgment and the technology is helping support that.
- [08:38]
A
Yeah, absolutely. You don't want to introduce new problems when you're trying to solve challenges. But especially with any emerging technology, those risks definitely exist. So walk me through some of the solutions that AID has developed over the past few months. Whether it Comes to preparation, response, immediate response, and then recovery.
- [08:58]
B
So when I talk about. Real quick, though, your, your last point about introducing risk into technology, I just, I think it's important to note this is not a new concept. Right. Every time we bring any sort of technology into emergency management, you're thinking about risk. Because by default, emergency management, at least in, you know, in most of the phases, is about risk management. You're thinking about what actions can I take to make the problem less in preparedness, what can I do to be, you know, more ready? So I reduce my risk in response, how do I move resources faster to reduce the risk and then recovery, how do I clean up faster to stabilize and reduce risk overall? So in the technology space, it's not new. Think back maybe 20 years ago, we used to use what were called T cards, and they were little resource cards where you would track an individual capability. A fire truck goes here, an ambulance goes there. Now that's all advanced through technology now. And you have blue force trackers and these great situational awareness tools. Understand that there's, you know, the systems are only as good as the data that you're putting into it and the availability to use it. So one of the things that we've been talking about with all of the practitioners and emergency management community is these tools are amazing and the advances in technology is fantastic. Just like any tool in your toolbox, remember to have a, we call it pace planning, but a primary, an alternate contingency and emergency plan. So if it doesn't work as you expected, you can still meet your mission requirements. So risk calculation and understanding the technology is not new. It's just the speed in which a lot of the AI tools are being developed and implemented is much faster than we're used to overall for technology advancement.
- [10:35]
A
Yeah, absolutely. I think that's, I mean, thank you so much for that, that history lesson as well. Again, before we go into the solutions. Yes, because AI and these, all of these tools are kind of advancing a lot faster than a lot of previous technologies. Can you think back over the past couple decades, if there were any new tools for the emergency management space that introduced a level of advancement or efficiency in addition to the same level of risk and kind of like a learning curve that AI has?
- [11:06]
B
Yeah, I can offer a couple. I think probably the one that resonates the most is the advancement of geospatial Information systems, or gis. You know, you think back to early disasters, you'd respond to a disaster location, you pull out a ADC map, you know, grid map and sort of start to figure out how you do damage assessments. And you'd drive down a road and you'd mark it on a, on a printout of, you know, damaged, destroyed, impacted and those kinds of things. And you'd collect these damage assessments, they'd roll up into a narrative report and you would start to determine where you would maybe push some, some resources towards a particular area. Right. With the advancements, and I'm, I'm using a more simple example, but with the advancements of geospatial information systems where you can actually do aerial imagery collected by satellites or drones or piloted aircraft, the technology is so good in some of the imagery now that you know immediately when the situation allows, meaning if a storm passes or smoke clears from a fire, you can go and collect this amazing amount of imagery and do damage assessments so quickly that you can do immediate payouts to individuals, you know, private insurance, public assistance or individual assistance. So, so much advancement in that space. The other one I would say is, is in and around dispatching. So your basic 911 dispatcher, right. The old school was a call came in and if you've ever seen like a, either old video or you're moving the lines and plugging things in, then you went to these cards where it was a run card and you said, okay, hey, this call is at, you know, 123 Main Street. I'm going to send these things. Now it's all automated. The call taker can focus on the person they're talking to. Where's your emergency? That's the first question they're going to ask, not what your emergency is. Where is your emergency so they can start sending resources. So I think those couple examples really lend some background to how much we're advancing. Again, remembering that we're adopting technology appropriately and trying to figure out where it's going to make, make the, the emergency manager better and more effective. And it's not always the thing that's the most pressing because if you can offset basic admin time, drafting, email writing, reviewing plans, developing an exercise that take hours or days can take minutes. Now, I think all of these advances allow people the extra time to focus on the tasks that either require human, human contact or, or human judgment.
- [13:28]
A
Yeah, absolutely. Especially as you mentioned, that a lot of these emergency management offices, they're operating with really limited personnel. So any time that can be freed up is a game changer for them, I imagine.
- [13:40]
B
Absolutely. And it's really about not just freeing up your time, but offsetting some of the administrative tasks that are so burdensome to get the person back into the thinking process. The training, the drilling, the exercising. On the preparedness side, the ability to get out during disaster and interact with the survivors directly. And then on the recovery side, taking, because this is important, a minute to decompress, write your notes, draft an after action report and make sure that you're going back through that process.
- [14:09]
A
All right, so walk me through some of the tools, some of the solutions, some of the technologies that have emerged.
- [14:16]
B
So right now where we are in the process, we've finished our research phase and as we transition into this more action oriented phase, a couple of things. So one is figuring out how we can disseminate this information appropriately and usefully. Emergency managers, as we talked about earlier, the time is not a luxury that a lot of people have capacity to read, digest and then act on. The information is also a critical shortage. Not because there's not interest, it's just how do I do it? So one of the things that we're endeavoring to do is writing out some roadmap, some technology adoption rules of the road, if you will, about how jurisdictions can go about looking at not just a specific tool, but an approach overall to AI adoption. Some jurisdictions are super advanced in this process. We want to see what we can do to bolster and help. They've already done. And some areas genuine interest, but no real ability to understand where to start. So trying to develop some products that will be helpful for each of these jurisdictions and each of the people that want to advance their AI adoption overall,
- [15:27]
A
do you find that? I mean, obviously natural disasters are increasing all over the country, but there are regions that experience more specific disasters than others. I'm thinking of, you know, Florida with hurricanes, California with wildfires, Louisiana with again hurricanes and flooding, especially Texas with flooding. Do you think that there are specific tools that are more useful or tailored towards specific disasters, or can some of these be applied in a more general way to assist emergency managers with admin work, situational awareness, streamlining efficiencies within the process or their operations to just give them back time?
- [16:07]
B
That's a great question, and the answer is both. So there are a variety of different tools that are specific to an individual hazard. So a couple of examples. There's tools around fire starts. So infrared radar capability that looks at large areas of land, you know, out in Midwest or in Texas or down southwest Midwest areas that aren't highly populated but tend to have initial fire starts. If that firestart can be identified through, through a capability that's doing monitoring, we've even seen some technology where autonomous vehicles can go up and deploy fire retardant or in certain cases other chemicals to stop that fire before it spreads anywhere is an amazing technology, but very specific to wildfire. Another example is stream gauges in and around heavy flood area. Deploying multiple stream gauges and then having the technology to map out where you think that that water is going to move downstream to help identify one, obviously the water intrusion, but might be able to help inform evacuation guidance for those jurisdictions. Those are very specific examples of where maybe it's a firestart technology or a flood technology, but then there's other capability that's broader like we talked about situational awareness tools, exercise design development tools, grant review products, checklists for special events. So those are more incident agnostic in the sense that the threat itself doesn't matter, but they're applicable to all the different phases of emergency management.
- [17:46]
A
That makes a lot of sense to me. I mean over the past few months I'm, I'm curious, how receptive are these jurisdictions, these entities, these nonprofits, these groups, how receptive are they to adopting these new technologies? I know change is hard. The procurement process, especially with government, can be, can be long and training can be, you know, a bit burdensome. Especially if you're not familiar with using some of these tools right away and in emergencies you need to be able to use them immediately and effectively. How receptive and interested are these places in wanting to adopt and use these new tools?
- [18:21]
B
So this is one of the most appealing parts of the AID initiative and really one that, that drew me to want to work on it as well. It is a nonpartisan, non vendor based, nonprofit approach to examining AI adoption and emergency management. So an honest broker, if you will, in the space between not just emergency management practitioners, but all of the academic research that's going on in universities and it's, it's vast and super impressive. And then the technology industry that you have both large hyperscaler companies, regular tech companies that are in existence, and then even other non tech, but private sector entities that are pushing or using AI adoption. So the 8 initiative is designed to be that, that large tent to bring everybody in. If you've been to like a tech conference, it's fantastically detailed around tech. When you go to emergency management conference, it's amazingly detailed around em academic conferences. But you know, when you can bring together all these people then plug in voluntary organizations into their group. It's our, so far what we've seen is people are very receptive to the discussion because we're not selling a particular product, we're not providing individual vendor, but rather saying, one, these are our observations. But two, talk to us, inform us and help us sort of de risk your thinking. An emergency Manager can spend 10 or 15 minutes on the phone with our team and spout out what their concerns are, and then hopefully what we're generating is a much more useful return to be able to say, if these are your challenges, we could recommend this type of tool or this type of approach. So I think, to answer your question directly, people are very receptive to the idea of technology being adopted in emergency management. Tech companies are, for very good reasons wanting to push into this space. There's a need on one side, a capability on the other. How do we be that that entity to bring everyone together and have the conversation around what's best? Because not every tool is going to work for every person. So it's more about the approach than the tool.
- [20:22]
A
Absolutely. I mean, it sounds like what I'm hearing is you're meeting people where they are meeting people where they are in terms of their training, their use cases, tailoring these, these tools and your recommendations to really fit the mold of what their region and their emergency managers can, can handle. Is there any type of accountability or ethical frameworks? I know you mentioned wanting to adopt some of these policies to make sure that the jurisdictions and the emergency managers and the residents that they're serving feel comfortable using these tools. What are you recommending to them in terms of accountability frameworks or ethical guidelines in case something does go wrong?
- [20:59]
B
Is a great question and one that I think goes far beyond the emergency management community. Even today there was the signing of the latest executive order on AI overall and a large national international discussion about governance within AI systems. What are you relying on to make decisions? Where's the human in the loop? So a couple of, of key themes for us. So one is understanding truly that technology and the data that is in it is an excellent first indicator or last validator. It's not absolute, right. Just because your system tells you something, don't take that as default. And we know that, but you often have to say that part out loud so people know that it's there. The second is you have a human in the loop. And the example that I use when I talk about this regularly is, and you may have even done this today, you wake up, you look at the weather app, right? And you still question it. You say, okay, well, maybe the weather says it's not going to rain, but I'm going to throw a rain jacket in my bag anyway. Or you get in the car and you pull up Waze or Google Maps and you plug and address it and you say yeah it says that but I know a better way. Right, so that's, that's the perfect example. Maybe again over simplified, but a perfect example of where the human is still in the loop. So using technology and data to inform your decision making is a critical point. And then the third issue for us is making sure that we are recommending emergency managers that they're transparent about what computer systems and technology systems and AI tools they're using in their space. I think there's a, on both sides there's an over concern about or stigma about, well did you use AI to generate this? Because the kid is amazing, so we should understand it. And conversely, just talking about how this information is generated using technology, you know, forecast, predictive analytics that we rely on as citizens every single day and our managers rely on for hazards and threats. Huge adoption of AI capability and we love it because it allows us to look around the corner, right? Look through that 72 hour window to figure out what those forecasts really are. Urban search and rescue capabilities that use computer visioning and drone technology that allows us to find people faster when they're, they're buried or in a large area. Like all these things that we love and embrace, we just have to think about it from a perspective of being clear and transparent about what tools we're using for which jobs. For those who are interested in learning more, aidinitiative.org is the website for this initiative. Overall there's a intake site there that I will shamelessly plug. If anybody wants to find out more about the initiative has questions. If maybe some of the discussion this podcast sparked an idea you want to provide input into what we're working on, Please feel free to reach out through the aid initiative.org and I would just say as we continue down this road that this as mentioned, the speed in which the technology is advancing is vast. We know that there's demand signals on the emergency management side. We're looking for every avenue that we can to help marry that up and minimize risk to those who are impact.
- [24:01]
A
Thank you to Jeremy Greenberg for participating in that conversation. You can subscribe to Priorities Podcast at Priorities Podcast and wherever you get your podcast. While you're there, be sure to leave a review or rating on the podcast page. That small extra step helps more people like you find the show. This podcast is a production of Scoop News group in Washington D.C. adam Butler and Carlin Fisher help put it together. Until next week, I'm Sophia Foxo. Thanks for listening.