Imagine if in 2015 we had the ability to estimate how big the refugee wave from the war in Syria would be? Or to predict the effects of the 2008 economic crisis, which is often cited as one of the greatest contributors to undermining trust in institutions in modern history. Or to forecast the likelihood of social polarisation intensifying and boiling over into violence, as it did in the assassination attempt on the Slovak Prime Minister Robert Fico in May 2024.
This would allow us to be MUCH better prepared for social crises. It would also enable us to test the impact of public policies. For example, what happens if we raise VAT in Slovakia by 3%, and at the same time make the criminal law laxer, both in an environment with very high societal polarization?
Such analytical predictions are rather difficult to make accurately. The world is complex, and even to simulate physical events requires a lot of data and knowledge. But it is even harder to simulate society. Gravity does not change its opinion and alter the strength of attraction of celestial bodies because it had a „bad day“. People do.
But there are patterns in human behaviour. We need to eat and sleep, and we change preferences according to health, age and wealth. Whether we live alone or with someone, and whether we have children, has a big influence. The degree of trust in other people or the feeling of belonging to a collective identity - e.g. "Slovak" or "Cassovian" - also matters a lot. We can get a pretty good sense of what we perceive as risk, and when we feel safe. Our ideas about morality also have a big influence on solidarity, political convictions and voting behaviour. All this can be measured.
Thus, simulating society is possible - just very complex. To do it, we would have to evaluate thousands of indicators for each individual across a population of many millions. Mastering that level of complexity has been beyond human capabilities - until now. The availability of artificial intelligence (AI) has changed everything. Only a few years ago, all the money in the world could not buy what a few people with the necessary skills can do today.
And just such people have gathered in Slovakia. Social scientists and programmers from the Institute of Ethnology and Social Anthropology of the Slovak Academy of Sciences, the DEKK Institute and the analytical company CulturePulse joined forces and decided to simulate the entire Slovak society - literally five and a half million artificial intelligences, one for each inhabitant.
This method is known as MAAI - Multi-Agent Artificial Intelligence. So far, this method has been used to simulate a variety of complex systems, from ocean currents to global logistics chains, and in recent years the first digital twins of large cities have been created – such as London and Venice. Simulating entire societies is therefore the real new challenge.
These MAAI models are filled with psychologically realistic simulated agents in sociologically realistic relationships and connections. They go to work, get married, shop, watch TV and read newspapers, get happy and get angry, worry about the war in Ukraine and are unhappy about the rise of cost of living. The digital twin of the Slovak society will thus make it possible to simulate social dynamics with a high degree of realism.
The most important issue is - how accurate will the digital twin be. We assume it will be accurate for macro-analytical purposes, but we will know for sure in 2027, when it's put through its paces. But based on experience, we think it will be accurate enough to enable a revolution in the design and testing of the impact of public policies and the prediction of certain types of crises, such as social unrest or declining social cohesion.
We think so because we have run similar, albeit simpler, simulations in the past. The DEKK team, in collaboration with CulturePulse, who did the analytical heavy-lifting and programming, carried out a simulation of population relations in Bosnia and Herzegovina for the UNDP. We have also collaborated on a similar conflict modelling exercise in Northern Ireland. CulturePulse even worked on a model of Israeli-Palestinian relations just prior to the 7 October 2023 Hamas attack, which indicated a high risk of conflict in the coming months, despite many media commentators talking more of a reduction in tensions. We therefore think that these models can be quite accurate.
A working prototype of the digital twin of Slovak society will be available to scientists from the research team in 2027. If we succeed, it will be the most advanced macro-analytical tool in the history of Slovakia.
The acquired know-how will enable us to improve the simulations later on, and at least technologically it should be that hard to upscale to simulating the whole of Europe or the whole world later on.
If you have read Isaac Asimov's Foundation, then science fiction is just becoming a reality.
But it also legitimately raises concerns. AI technology is highly exploitable - it's the wild dream of dictatorial regimes or criminal organizations around the world. But in general, with highly available technology (as opposed to, say, nuclear technology), it makes no sense to impede one's own progress, because whoever wants to abuse AI is already doing it. Dictatorial regimes have already deployed it to monitor their populations, and criminals have already set up more sophisticated illegal e-shops in the Dark Net or complex AI-powered money laundering schemes. This is not something we can prevent in academia. What we can do is look for positive uses for the widely available modern technologies, both to serve ordinary people and to make democracies more effective. Making better tailored public policies and predicting and preventing crises is what we want to contribute.
Let us keep our fingers crossed.
Sources:
Bauer, P. "Digital Twins of the Earth with and for Humans." Nature 607, no. 7920 (2022): 655–660. https://doi.org/10.1038/s41586-022-04928-2.
Burken, Ch.B. at.al. "Agent-Based Modelling of Values: The Case of Value Sensitive Design for Refugee Logistics. " Journal of Artificial Societies and Social Simulation 23(4) 6, (2020). http://jasss.soc.surrey.ac.uk/23/4/6.html
McGibney, Alan, Susan McKeever, and Dympna O’Sullivan. "Human Digital Twin: A Survey." Journal of Ambient Intelligence and Humanized Computing 15, no. 2 (February 2024): 765–787. https://doi.org/10.1007/s12652-023-04691-w.
O´Lone, K., Gantley, M., Lane, J.E., Shults, F.L. "Employing Multi-Agent AI to Model Conflict and Cooperation in Northern Ireland." New England Journal of Public Policy, Volume 36, Issue 1: The Changing Character of War and Peacemaking.https://scholarworks.umb.edu/nejpp/vol36/iss1/6/
Shults, F.L. "Simulation, Science, and Stakeholders: Challenges and Opportunities for Modelling Solutions to Societal Problems." Complexity, Volume 2023, Article ID 1375004, 10 pages https://doi.org/10.1155/2023/1375004.
Shults, F.L. et. al. "A Generative Model of the Mutual Escalation of Anxiety Between Religious Groups." Journal of Artificial Societies and Social Simulation 21(4) 7, 2018. http://jasss.soc.surrey.ac.uk/21/4/7.html
Shults, F.L., Lane, J., O´Lone, K. "Simulating Sustainable Societies: Uncertainty, Complexity and Multi-Agent Artificial Intelligence Modelling." 2023/2024 UNDP Human Development Report, Background Paper no. 1-2024. https://leronshultsblog.com/wp-content/uploads/2024/12/culturepulse-2024-undp-hdr-paper-simulating-sustainable-societies-1477638360-e1733814730822.pdf.