Agent based modelling – what is it and what can it do for us?

Our latest Hutton Highlights podcast takes a delve into the world of computer simulation. We recorded this episode during the European Social Simulation Association Summer School in Social Simulation. The week-long event at the end of August saw people from all over the world come to our at our Craigiebuckler campus in Aberdeen to look at how an advanced method of computer simulation called agent-based modelling can be used to tackle problems from climate change and energy to health and rewilding. 

Interviewer:

Elaine Maslin, Media Officer 

Guests: 

Gary Polhill, senior research scientist at The James Hutton Institute and lead organiser of the European Social Simulation Association Summer School.

Connor Lovell, an ecologist doing a PhD into rewilding and ecosystem processes at the institute of Zoology at King’s College London.

Mariëlle Rietkerk from Delft University of Technology in the Netherlands. Marielle is doing a PhD in the energy transition. 

Dr Anu Mishra is working for the Bill and Melinda Gates Foundation in the US in the Institute for Disease Modelling as a senior research scientist. She has a background as a biostatistition. 

Ryu Koide is a senior researcher at the National Institute for Environmental Studies. 

For more information about what you’ve heard in this podcast, visit: 

European Social Simulation Association Summer School in Social Simulation 2023 | The James Hutton Institute

Exascale computing could supercharge crisis response capability | The James Hutton Institute

Don’t forget to visit us at www.hutton.ac.uk 

Thanks for listening, we hope you enjoyed this glimpse into our world.

We look forward to bringing you more insight the world across food, energy and environmental security in future episodes of the Hutton Highlights podcast.

Remember to subscribe and follow us on Facebook, Twitter, Instagram and LinkedIn.

Transcript

Unknown
Welcome to the latest Hutton Highlights a podcast from the James Hutton Institute. We're here to bring you a glimpse into our world leading research across food, energy and environmental security. For this episode, we're here on sites that are Craigiebuckler campus. We're in the midst of the European Social Simulation Association Summer School. It's a real hive of activity with people from every corner of the globe, from Norway to Canada to Saudi Arabia and Japan.

00:00:35:00 - 00:01:05:08
Unknown
But first, I want to set the scene. There are a lot of problems in the world where there isn't a simple way to find an answer. You just can't add one plus two plus four equals seven. They're just not that simple. We're talking about behavioral challenges, such as what drives people to make certain choices in certain circumstances. These kinds of choices can have major global implications, such as on climate change, health outcomes, or how we use and distribute energy, and whether as a result of that, our energy grids collapse or not.

00:01:05:10 - 00:01:32:18
Unknown
Traditional computational methods don't always offer a suitable way to investigate some of these challenges, which is why increasingly a method of simulation called agent based modeling is being used. Especially with social interactions are involved. It's being used to simulate things like what happens when we change policies during a pandemic or what happens to energy grids when large enterprises ramp up or down their energy use or there's a war knocking out supplies here in Scotland, a specific uses around land use.

00:01:32:21 - 00:01:54:22
Unknown
How we use land, how changes to how we use land will impact it and the communities and society around it. It's a fascinating multidisciplinary method that has brought a large number of experts from ecologists and psychologists to environmentalists here together in Aberdeen, to talk about agent based modelling and how it can help them to solve the challenges that they're facing in their fields.

00:01:54:24 - 00:02:22:01
Unknown
Event organizer Dr. Gary Polhill is here with me now and can tell us more. Maybe start with what agent based modelling is. Okay. So put simply agent based modelling is a kind of computer simulation in which individual people, businesses and sometimes even governments are explicitly represented. And we also show how these social entities interact, in effect, with each other.

00:02:22:04 - 00:02:49:15
Unknown
And sometimes even how they are influenced by and influence the environment and the space around them. Okay, That's interesting. And is it a new technique? How different is it to other techniques that have been used? It's it's not that new, actually. So the one of the first agent based models, I think, was a model of social segregation, which was done in the late sixties and early seventies by two researchers, Skoda and Schelling.

00:02:49:17 - 00:03:15:03
Unknown
What they showed was that with a very small preference for living near people like you, you could end up if people were free to move wherever they wanted with a segregated society. But in terms of its difference with other methods of computation, computational modelling, the main difference, I'd say, is the explicit representation of individuals and the effects that they have on each other.

00:03:15:05 - 00:03:36:24
Unknown
This tends not to happen in alternative methods. It started to be older. It came first use in the sixties. What's happened since then, and I understand this bit of a watershed during the pandemic and it's increasingly been useful. For a long time. The early models were quite theoretical in nature. So the shedding and Skoda model of social segregation was very simplified.

00:03:36:24 - 00:04:06:22
Unknown
You couldn't really simulate any real society with it. It just explained how with a simple preference, you could end up with segregated societies. But of course, the the factors influencing where you move are diverse. You know, like not least the affordability and attractiveness of the site and the schools nearby and stuff like that. So for a long time we had these more theoretical models, but then perhaps maybe 15 to 20 years ago, we started working on more empirical models where we'd be working with real data.

00:04:06:24 - 00:04:29:13
Unknown
And this introduced some real challenges for us as a community. But the real watershed, as you said, happened in the in the COVID crisis, although they'd had been empirical models before then, it was during the COVID crisis where, of course the movements of individual people and the way they interact with each other and affect each other is critical for the transmission of the disease, like COVID.

00:04:29:15 - 00:05:02:12
Unknown
So you could explore scenarios that would restrict people's movement and see how they affected the movement of the disease. And they were used by various governments, local and regional, throughout the world. And until now or until more recently. And I guess because of that, it had been a bit more of a restricted discipline. But I hear that there's often just one nature based model or in an institute, and it's not it's not something that's been that widespread, but it seems to be gaining a bit more traction now because of the potential that's been seen through this modeling that was done in the pandemic.

00:05:02:13 - 00:05:23:19
Unknown
Sure. And also, there's a lot of work being done by various researchers. I mean, they may be on their own in in very small teams in universities, but there's been a lot of work, particularly the University of Surrey, for example, where they work very intensively with policymakers and started to get agent based modeling being used more in policy circles.

00:05:23:21 - 00:05:49:06
Unknown
But yes, you're right, And often agent based modelers are the only person in the department, possibly the only person in their university working in the field. So the European Social Simulation Association is a really vital lifeline for them to engage with and interact with other members of the community. Some of the people coming to the summer school today, their motivation in coming primarily was to meet other agent based modelers.

00:05:49:08 - 00:06:10:24
Unknown
Thank you, Gary. I think we should now have a chat with some of the participants to see what sort of problems that they're looking at with agent based modeling. So that I've got Connor here. Connor is an ecologist, so not a computer scientist. He's doing a PhD into rewilding and ecosystem processes at King's College, London and London Zoo.

00:06:11:01 - 00:06:43:02
Unknown
It's hugely topical subjects here in Scotland. Can you tell us more about your research? Yes. Hi, my name's Connor. I'm a second year PhD student working on rewilding in Scotland. The focus of a lot of my research is on the communities, on the essentially the case studies I work on, which consists of the species, but also while both trying to better understand how those species impact the ecosystem through ecosystem processes, things like grazing and primary productivity, and then how those carry on impact ecosystem functioning.

00:06:43:04 - 00:07:16:03
Unknown
Yeah, rewilding is quite a big topic in Scotland. Can you give us a sense of what is rewilding about as a rewilding is a more sort of novel and new nature conservation approach where rather than trying to restore a particular particular area of land to a particular habitat, say a woodland or a marshland or a grassland or a wetland, and we instead try and restore ecosystem processes and ecosystem functioning, with the idea being that those ecosystems and the more biodiverse and more resilient sort of future disturbances such as climate change.

00:07:16:05 - 00:07:39:00
Unknown
Could you give an example of that? So I think people will have heard about links being introduced and things like that. So. Well, there's different levels of what can be involved. Yeah, exactly. So I've not done any work really on predator reintroduction at the moment. But and so, for example, the wild boar have on the site I work on, they perform a process known as rooting, which is where the root of the soil.

00:07:39:00 - 00:08:01:19
Unknown
And the idea of this is that that sort of moderate, low level disturbance improves biodiversity by opening up. So breaking or homogenous plant communities and opening up the soil so new plants sprout there, it's more biodiverse. So it can be something quite, quite complex. And I understand that you're working with Gary here on how to use agent based modeling for your research.

00:08:01:21 - 00:08:28:02
Unknown
So could you tell us how agent based modeling could support this work and why you've had to turn to this type of modeling instead of using other simulation methods? Well, what's potentially just because I've never done agent based modeling before, I had a hint that maybe this will be useful because rewilding itself is obviously situated within sort of a social ecological system in the way that it's all and it's all well and good.

00:08:28:02 - 00:08:59:00
Unknown
Me as an ecologist going, Oh yeah, let's change this language to 12 string produced by a ball or everything, but if we don't appreciate that within the framework of a social ecological system than normal, those interventions will work or be implemented anyway. So, you know, there might be a pose that they might not work on people's communities. And this is what I think I'm hoping anyway, an agent based modeling might come in handy because it seems to be very good at linking sort of these ecological systems to these more social systems.

00:08:59:02 - 00:09:20:10
Unknown
Yeah, fantastic. What's what's been the most interesting thing that you've learned so far throughout this week? So we did something called causally feedbacks sort of modeling, which is where essentially you take a thing. So for example, in relation to my work that might be population size of wild boar and you think about the causes of this not population size and the consequences.

00:09:20:10 - 00:09:43:24
Unknown
So for example, something that would obviously increase all levels might be fall birthrate and something that might decrease both levels would be death rates and things like that. But then you want to tie in how you want to tie in, how the consequences of salt ball levels, so say biodiversity improvements that feeds back into the causes. I find that quite interesting.

00:09:44:01 - 00:10:12:24
Unknown
Mainly because it allowed me to sort of try and start trying to start trying to picture my own work in that framework. But also it was interesting thinking back to what I've done work before and I've sort of inadvertently drawn these diagrams without even realizing what I was doing. I think quite interesting using agent based modeling with that either enabled to enable you to do things that you couldn't do before, or does it open up new areas that sort of expand the possibilities of your research?

00:10:13:01 - 00:10:34:03
Unknown
I think it will put me in good stead to analyze these social ecological systems where you have the social side of things, the cultural side of things and people's norms and how people like to do things and how they make that money, but also then tied in with the ecological side of things for the fact that, like wild boar and modern levels will likely improve in spite adversity.

00:10:34:03 - 00:10:50:12
Unknown
But that might then tie into conflicts with people might not want borrowing their lands because they turn off their lawns. So I think agent based modeling will be quite good to tie all those different aspects and that sort of system together, you know, bringing different sciences together. Yeah, right. Thank you very much. Thanks.

00:10:54:02 - 00:10:59:04
Unknown
Next up is Mariella Kerk from Delft University of Technology in the Netherlands.

00:10:59:06 - 00:11:25:21
Unknown
Mariella is also undertaking a PhD. This is into the energy transition. Mariella, tell us a bit more about what you are looking into. Yeah, so I'm a social psychologist and I'm working looking into the energy transition, as you just said, and I am especially looking into the integration of behavioral science and behavioral methodology into agent based models that regard the energy transition.

00:11:25:23 - 00:11:47:24
Unknown
There are a lot of these behaviors. So the buying the expensive stuff and the more habitual ones and one of the obituary ones is changing energy use or shifting it to another time of the day. And when wind and solar energy is available. And that also has an additional problem because everybody does that at the same time, the grid is getting overloaded.

00:11:48:01 - 00:12:11:11
Unknown
And my research is about trying to prevent that. So as a behavioral researcher, I'm going to trying to nudge people into using appliances to not at a different time of the day, but not at a time that everybody assumes everybody else is using it already. Yeah, So kind of balancing out that grid system as we get more renewables and intermittent times of energies and energy.

00:12:11:11 - 00:12:43:03
Unknown
Energy use and grid systems are hugely complicated. So it makes sense that you've turned to something like agent based modeling because yeah, how about social behavioral element? How did you come across it and how do you hope it will happen? Well, to answer the last one, I, I worked as a social scientist for a few years now and I I've learned that policymakers use models and scenarios a lot in aiding their decision making, for instance, when they're deciding on the height of a subsidy, for instance, for electric cars.

00:12:43:05 - 00:13:11:16
Unknown
And what happens with these models that they are the agents in these models are rational decision makers, so they don't see their environment very much. And it's far more complicated than the models they use nowadays. So I hope agent based model will be a tool that I can use in my research to get some complexity into that and to have a more realistic model which I can say predict, but which which makes the dynamics more visible.

00:13:11:22 - 00:13:47:24
Unknown
Yeah, Teases out some more of that behavioral social complexity. Yes. Yeah, yeah, yeah. And how did you come across it? What did you. And as a good question, the goal was I don't know agent based modeling for a long now because I didn't know that it existed. I was used to scenario of buildings, which is something different and actually I, I started to look into it when it was in my job description and I thought it was really nice because it has one agent making a decision and interacting with other agents, which is, well, totally different from the way it is used now in the Netherlands.

00:13:48:01 - 00:14:18:06
Unknown
What's been most interesting thing that you've learned this week and how how might help what you're doing? Well, I'm just getting into being a coder, which is difficult enough already because it's I think it's really complicated. I think what's really interesting is that you see the whole process from developing, developing a model from beginning to end. So it's not only about writing a code, but it's also about implementing a theory into it and and getting data to it.

00:14:18:06 - 00:14:35:16
Unknown
And I think I had some feel about the process, but really living it through and simulating it actually in the classroom is really helpful to see what happens. Yeah, yeah. And certainly been a lot of buzz and we can hear the chat even though it's lunch break, people are enthused and really interested in talking about this. Thank you.

00:14:35:19 - 00:14:37:13
Unknown
Thank you very much. Thank you

00:14:41:02 - 00:15:01:06
Unknown
So we've come from rewilding and ecosystem processes to energy grids. Our next guest is another completely different area, and that's public health. Dr. ANU Mishra is working for the Bill and Melinda Gates Foundation in Seattle in the US, as a senior research scientist in the Institute for Disease Modeling. She's also an honorary research assistant at Imperial College London.

00:15:01:08 - 00:15:38:07
Unknown
She has a background as a biostatistician. And tell me a bit more about your research. Yeah, so I work on maternal trying to model maternal, newborn and child health systems, whether that be like a hospital system in specific countries or maybe like a disease process, like something related to maternal mortality or child mortality. And we like to focus on low and middle income countries or developing countries where, you know, sort of resources might be hard to access or there might be challenging problems about putting in new interventions, you know, kind of into these systems and how how widespread is that?

00:15:38:07 - 00:15:58:11
Unknown
What kind of like the scale of the kind of challenges that you're looking at? Yeah, sometimes it can be as small as, you know, can we put an intervention in a hospital that might help, you know, prevent maternal mortality? Or sometimes it's as large as how do we implement an immunization program in the whole of Nigeria. So the scale geographically can vary quite a bit.

00:15:58:15 - 00:16:23:08
Unknown
Yeah, well, that's quite complex. Yeah, because the idea is that the sort of things that you're working on at the moment. Yeah. Right now really thinking about a lot about maternal mortality. So kind of mapping out the processes of maternal mortality and trying to understand what are the different levers or focusing in on Nigeria and, and trying to understand what are the different levers we can pull to sort of think about how you can reduce the the burden of maternal mortality.

00:16:23:10 - 00:16:49:05
Unknown
They're probably going to have quite a lot of complex things going on, social and behaviors and, yes, differing in different countries as well, and even maybe different different hospitals. So would what are the exact sort of problems that you're looking to address using agent based modeling? Yes, I think you hit the nail right on the head. So like my background is in statistics, but when thinking about these models and statistics, it's we're just in a very complex setting.

00:16:49:05 - 00:17:22:14
Unknown
We're thinking about things changing over time and space. We're thinking about, you know, the heterogeneity of a lot of these mothers and women who are going to hospital. They might be quite different, their experiences might be quite different. And also, you know, in statistics, we need a lot of data to help do these models. But now we're working in settings where we don't have a lot of data either because we're trying something new, like a new in a clinical or public health intervention that we haven't collected data on or, you know, they're just in places where there isn't a lot of data collection going on.

00:17:22:14 - 00:17:48:24
Unknown
So kind of both the complexity of the system itself and the lack of data that we can use statistically. So hoping agent based models can solve that problem. Yeah, it sounds interesting. So from what I'm hearing as well, there's quite a big crossover that age based modeling can do in terms of individual behaviors, like an individual mother's interactions and how she works, but also then clinical settings and how clinical settings change.

00:17:48:24 - 00:18:13:18
Unknown
So you've got very to two different systems happening there that can work together or against each other and trying to unpick that and understand it. Yeah, absolutely. And then, you know, with the type of work I do kind of relating that back to policy, so going sort of from that individual level as a mother all the way up to what's the right policy implications, you know, maybe on a state or national level in a country.

00:18:13:20 - 00:18:34:14
Unknown
And how much have you done with agent based modeling today? Is there something new that you're learning about? Yeah, this is quite new for me. So some of my colleagues at the Bill and Melinda Gates Foundation have used agent based modeling in their work to kind of model some other things related to women's reproductive health. But this is the first time, you know, where I'm thinking about it in this maternal, newborn and child health space.

00:18:34:16 - 00:19:02:02
Unknown
I'm quite excited. Yeah. Yeah, right. And what's been the most interesting thing you've learned this week about agent based modeling and what it could do for the work you're doing? I think one of the most exciting, like I'm truly excited about is how inherent space and time are to agent based models. So when thinking about a lot of the problems I've work on, as I mentioned, the geographical scale can be quite big, quite small, and it's and it's very complex and interacting with time.

00:19:02:04 - 00:19:27:20
Unknown
So being able to model that space inherently is quite exciting and it will come up in things like when we think about like mothers trying to get to the hospital, we want to account for things like roads might not be that accessible or there might be like a large desert or not a lot of hospitals nearby. So being able to build that right into the model to help us sort of really think about how we're going to distribute an intervention is is really nice.

00:19:28:01 - 00:19:30:01
Unknown
Yeah. Thanks. Thanks very much.

00:19:33:17 - 00:19:55:12
Unknown
We're jumping now to another completely different topic where agent based modeling can play a role. That's the circular economy. This is being looked at by quite a senior researcher at the National Institute for Environmental Studies in Japan. Will you tell me a bit more about your research and what you're looking into? Yes. So we're doing the consumer behavior modeling in the circular economy.

00:19:55:14 - 00:20:25:15
Unknown
So, you know, the circular economy is about reducing waste and minimizing environmental impact through utilizing, you know, different losses of the product in society. And we have quite many strategies like selling the product, renting all the products or repairmen to reusing. There are plenty of strategies. But a tricky thing is the, you know, introducing these circular solutions or circular business models might not be always environmentally sustainable.

00:20:25:17 - 00:20:56:11
Unknown
For example, like doing showering, the product might increase transport or maybe introducing reusing. Maybe you will buy more because it's cheaper. So this kind of are at, I would say, tricky part needs some more assessment about environmental consequences from these circular solutions. And also taking a look at consumer behavior that's very crucial points. Yeah, interesting. So how are you looking to use agent based modeling for your work around circular economy?

00:20:56:11 - 00:21:18:17
Unknown
What problems do you want it to help you solve? Oh, yes. So the thing is that there are many different methods already applied to evaluate circular economies like material for analysis, to quantify the amount of waste and amount of resources, input or lifecycle assessment, which is to basically to quantify environmental impact or other simulation method like discrete event simulations.

00:21:18:19 - 00:21:40:22
Unknown
But none of them basically focus on consumer behaviors and take into account the heterogeneous consumer behavior or dynamic consumer behaviors or well, maybe, maybe give us an example of that. So what? Yeah, well, it's kind of a scenario where this might help you understand those behaviors. Uh huh. Yeah. So, for example, taking a look at like a sharing service fusions.

00:21:40:24 - 00:22:09:05
Unknown
So we need to model the consumer behavior to understand how people use the product. Maybe how long do they use it or do they really properly return it. And, and we also calculate the environmental consequences from from these behaviors. Yeah, but from the traditional method other than AVM could not really model the determinants of the consumer behavior changes.

00:22:09:07 - 00:22:31:02
Unknown
Yeah. So the point is that, you know, consumer behavior are heterogenous. So some people use sharing service in a more sustainable way, but some other people may not. But in this method we just assume like average behavior or typical behavior. And to say that, okay, this sharing service is environmentally friendly or not, but we using agent based simulation, we can take a look at different consumer segments.

00:22:31:02 - 00:22:57:14
Unknown
So with this segment, maybe this kind of service may work in terms of sustainability for other segment, maybe not working in terms of sustainability. So what we want to look at is try to identify which circular solution works in terms of sustainability, but also diffusion potential. So we need solutions which consumer accept, majority of the consumer accept, but also sustainably being used.

00:22:57:16 - 00:23:17:01
Unknown
Yeah, yeah. Which could really help roll out these kinds of things. Like I think we hear about more and more community led schemes like people putting tools into an area where a whole community can go and use those tools. But what, what can make those better and more sustainable and yeah, help people adopt those kind of behaviors. Yeah, exactly.

00:23:17:01 - 00:23:48:09
Unknown
That could be one application area. Yes. Yeah. What's been the most interesting thing you've heard this week? So what have you learned so far? I mean, the agent based simulation is really interdisciplinary. So for example, in the seminar there was a workshop with participants from different disciplines. So we work with I'm more from engineering backgrounds, but I worked with are experts in business and finance and economics and sociologists, and we work together to build a model and also, you know, try to interpret the model results.

00:23:48:09 - 00:23:54:24
Unknown
And I think that's a quote, Asian simulation and and we're learning about that here. Yeah, Fantastic. Thank you very much.

00:23:58:15 - 00:24:11:17
Unknown
So I'm back with Gary. We've heard some really interesting projects there and the challenges that these delegates are looking at to use agent based modeling, it must be quite satisfying to see that they're really taking up this idea.

00:24:11:19 - 00:24:37:24
Unknown
But for you, what's the challenge that you would like to solve using this type of simulation? I mean, it's very hard to put my finger on one challenge, but I think I'd go more abstract and talk about social and ecological and technical complexity. All of the problems we're looking at, I think, have these features in common that, you know, you try and do one thing to solve one problem and then up pops another one.

00:24:38:01 - 00:24:59:00
Unknown
And so you need to do more in the way of systems thinking, thinking more broadly. COVID was a very good example. So yes, lockdown helped in terms of stopping the spread of the disease, but it also had some knock on effects on the education of young children or the welfare of people who were maybe stuck with an abusive partner or something like that.

00:24:59:02 - 00:25:18:04
Unknown
So it wasn't necessarily all good. So when we're modeling COVID, we need to think more broadly. And I think that's what interests me most. How do we do that? And that's the challenge I really want to address. And what are the challenges to doing that in kind of some of the details and exactly how agent based modeling can do that?

00:25:18:06 - 00:25:37:17
Unknown
So I think the main challenge we face in agent based modeling is what I would call trustworthy agent based modeling. How do we know that the model we've built is really telling us something useful about the scenario we want to explore? And sometimes that's about creating the right social context in which what the model is doing is understood by everybody.

00:25:37:19 - 00:26:06:01
Unknown
But also sometimes that's about fitting it to data. Now, because these systems are complex, it's actually really hard to use traditional methods and metrics for measuring how well a model is doing. So we need to develop more and we also need to look more at methods of working with the social scientists and mathematicians and other experts in multiple disciplines to try and understand better how well these models are telling us what to do now if we take that into the real world as well.

00:26:06:01 - 00:26:27:10
Unknown
How how do we match these models with the real world and what are we doing? Is it kind of effectively a scenario testing tool so we can like throw something at it and go, What happens if that's right? But then how do you know that the scenario you're looking at and the model has told you something useful? So what happens if we lock everybody down?

00:26:27:12 - 00:26:50:05
Unknown
Well, yes, maybe we stop the spread of COVID, but then what else happens? What we don't know. And if we've not got experience of that historically, then it's going to be very hard to fit any data to it. So how do we know that? What the models telling us happens is reasonable or it doesn't have to be right necessarily it at least reasonable, something we might suspect?

00:26:50:07 - 00:27:08:07
Unknown
And I think that's the real challenge. Yeah. So with you started in your projects using or testing how agent based modeling could be used with exascale computing. Now something that we're starting to work on is that going to help with that challenge? Because you can do I understand you can do more computations faster instead of taking months and months to do this.

00:27:08:07 - 00:27:37:14
Unknown
So yes, look at more scenarios. So an exascale computer is one that has the kind of power that you would now have with a billion laptops. So that's a huge amount of computing resource. And if we can get everything working right, this project is really just exploring the space of that rather than actually trying to do it. If we could get anything right, you could get the answer back from some of these scenario questions in a fraction of a second rather than a very long time, like several days or even a month.

00:27:37:20 - 00:28:04:08
Unknown
If you were using traditional computing ever mind your laptop, we could be talking the cloud or what we would call high performance computing, exascale computing, so that much faster, which allows us to explore scenarios more effectively. That kind of compute computational power also lets us explore multiple possible outcomes from the same scenario. So of course it's not necessarily the case that there's only one single outcome that might happen.

00:28:04:08 - 00:28:23:15
Unknown
If we locked everybody down. There could be lots of different ways that that could pan out. I mean, of course, once it's actually happened, there has been only one history. But as we look towards the future, there are many, many ways that things could play out. And exascale computing, I think, could really help us with that. Could that help develop and build trust in the models as well?

00:28:23:15 - 00:28:54:18
Unknown
Because you could potentially test things that you could then in a safe test environment, play it out in the real world, which then test against what the computer system and potentially yes, there I think the the in terms of the trustworthy nurse, one of the main benefits would be if we were able to build on existing work. There's some work going on in this in the agent based modeling community, particularly University of Delft, where they're trying to develop reusable building blocks.

00:28:54:22 - 00:29:19:05
Unknown
Now you can imagine that if you have a reusable building block of code that simulates a particular of a subcomponent of the system, very well indeed. You could trust that and you could then bring it together with other reusable building blocks to start exploring some of these influences and impacts. Uh, in the into the future. Yeah. Interesting. I think that's one of the rabbit holes to go down.

00:29:19:08 - 00:29:36:17
Unknown
This will put some links in the program notes for people to go and find out more information about this and as well to do the same as go. So that'll be great. That's been fascinating. Thank you very much. And my thanks to Gary O'Connor, Mariella Agnew and Rio for being our guests and telling us about the work that they've been doing here this week.

00:29:36:19 - 00:29:56:15
Unknown
Thank you also to the listeners for joining us and do like and share the podcasts and of course, subscribe to us on your podcast platform of Choice. Just search for the Hutton Highlights podcast.