Welcome to Snapshots of Social Science - a monthly newsletter where we bring you recent developments in the diverse fields of social science. Learn more about us here.
This month, we will look at Agent-based Modeling. While this is not exactly a subfield of research like the topics of collective intelligence, cultural evolution, and collective action that we have covered in past editions, it is an important tool for these subfields—it helps model the behavior of systems they study. Let us take an ant society as an example. One way to think about the ant society is that there is a queen ant who directs all the other ants to do their own particular duties. This would be a top-down approach to model the ant society. But, what if we model the ant society in a different way: as a group of multiple ants that interact with each other and their environment using certain simple rules? That is, they have no central authority directing them: the ants just follow other ants following some simple cues. This would be a bottom-up approach—exactly what agent-based models do. Agent-based models model a system using interacting agents, and observe the behavior of the entire system. As we can imagine by this general description, these models can be used in widely different areas of science and humanities. In this newsletter, we bring to you the use of agent-based models in fields like urban planning and the environment, COVID-19, business management, sociology, computer science, and biological sciences.
Before we begin, please consider subscribing to ICJS Social Science by clicking the button below! This will ensure that each edition of our newsletter makes its way to you every month. We would truly appreciate it if you spread the word about us as well - our impact grows with each additional reader.
Research Map
News from the field
(A) How to promote residents’ use of green space: An empirically grounded agent-based modeling approach (🇨🇳)
Humans need green spaces, and designing these green spaces is a challenge for policy-makers. To design a green space, policy-makers have to consider the diverse needs of individuals, their accessibility constraints, and balancing green spaces with rapid urbanization requirements. Is there any way to make their decisions easier? An agent-based approach can help.
In this paper, researchers used survey data from residents in Shanghai to model their individual decisions on green-space usage in a computer-based simulation. In the model, they incorporated the heterogeneous characteristics of these individuals, and accounted for some irrationality in their decision-making. The researchers simulated three different green-space policies and observed how the model behaved. They found that, among different factors, the resident needs for social interaction and public participation are most sensitive to policy. They also found that improving social interaction, commercial function and public participation is more efficient in the early stages of the policy development, while improving safety, security, environmental quality, and accessibility is more efficient in the later stages of the policy development.
This research can be taken in multiple different directions. The authors of the paper suggested that further research could incorporate social networks into the analysis—modeling resident decisions based on how their neighbors behave. Also, this model could be generalized to other types of policies that impact forms of public participation that are not necessarily related to green space use.
This paper uses Agent-based Models more generally to look at residents in cities and the satisfaction of their basic needs depending on the physical environment around them. 🇨🇴 🇮🇹
Researchers here use Agent-based Models to look at different ways to manage Cape Town’s food-energy-water resources under different climate scenarios. 🇺🇸 🇿🇦
This paper uses Agent-based Models to model’s farmer decision-making of opting for biogas facilities by considering factors involved in their decisions and barriers to entry. 🇨🇭 🇩🇪
When the COVID vaccine rolled out, policymakers in a lot of countries needed to establish a vaccination strategy. Who would the vaccine be given to first? At what rate? Was the production capacity of the vaccine large enough to support the strategy? And how long between the two doses should people wait? It is hard to answer these questions because there wasn’t enough time to conduct controlled research experiments, and a lot of the features of the pandemic were unprecedented.
In such cases, computer simulations can help. In this paper, researchers modeled different vaccination strategies and their impact on the spread of the virus using agent-based modeling. Each individual was an agent, and agent interactions in networks modeled the spread of the disease. The researchers varied the efficacy of the vaccine, rate at which people were vaccinated, and delay of the second doses.
The researchers found that a delayed second dose strategy for people under 65 performed consistently well under all the different vaccination rates they tested. Although this model relies on assumptions that may or may not be true, such models can be useful because of the number of different possibilities they can take into account, and their ability to predict based on changing parameters. It is a good proof-of-concept that agent-based models can be a useful tool for policy-making.
This paper models COVID spread using Agent-based Models in a small town, taking into account factors like testing, treatment, and vaccination options, and illnesses with symptoms similar to COVID-19. 🇺🇸 🇳🇱 🇮🇹
This paper too models COVID spread with an Agent-based Model, with the finding that an early lockdown and contact tracing using smartphones can completely suppress the spread of COVID if there are enough smartphone users. 🇧🇩
This paper makes an Agent-based Model of COVID spread taking into account changing behaviors of people based on other models of fear and contagion, and tries to come up with a strategy that could flatten the curve. 🇺🇸
(C) Agent-based modeling of participants' behaviors in an inter-sectoral groundwater market (🇮🇷)
When individuals take part in a market, they make decisions based on the information they have, and their own personal preferences. In such scenarios, it is hard to accurately model individuals in a market using an average prototype that assumes every individual in the market acts the same way. And that is where agent-based models come into play—they can account for different and varying characteristics of agents, and simulate their behaviors to understand the behavior of the system.
In this case, researchers use agent-based models to model the groundwater trading market in Iran, where both buyers and sellers place bids and are matched accordingly. Agents make choices based on their psychological characteristics, and social cues in the market (or, the information that they receive). Running simulations and validating them on 8 years worth of data led the researchers to study the trading patterns more thoroughly. For example, they found out that the implementation of a dynamic cap-and-trade policy increases the total net benefits of market participants by an average of 27% per year while reducing the region’s groundwater drawdown by 56 cm.
Importantly, the findings of this study are helpful for sustainable environment management. Modeling markets in this way can help researchers study emergent patterns based on individual preferences. Also, these models can account for a changing environment and agent characteristics, making it easier for researchers to study a phenomenon over time.
Read this paper to understand how agent-based models are used in economics and finance, and this paper to understand how they can be used in accounting research. 🇺🇸 🇬🇧 🇦🇹
Researchers in this article elaborate on an agent-based model to understand the interactions between renewable energy sources, consumers and thermal power plants in the European Continuous Intra-day (CID) market. 🇸🇪 🇧🇪
This agent-based model helps us model disruptions in the supply chain of rare earth metals, and their impacts on manufacturing of subsequent goods. 🇺🇸
(D) Agent-Based Modeling: an Underutilized Tool in Community Violence Research (🇺🇸)
It is hard to predict and deal with community violence problems because of the way in which they erupt. Modeling contagion, or the spread of these activities through peer or environmental influences, is usually the way in which they are studied. Because the influences can be different for different people, agent-based models could be very useful for studying community violence.
Researchers in this paper identify three prior studies that use agent-based models to study community violence. They also describe the possible intervention strategies and policies that could impact such violence. The authors argue that there is currently a lack of use of agent-based models in this line of research, and that if used, they could be combined with expert knowledge to be used as a decision-tool for policy-making.
Indeed, a lot of societal problems can benefit from agent-based models, which can incorporate both individual interactions and structural factors in the environment. Expert knowledge on individual agent behavior and rules for interpersonal interactions, combined with simulated outcomes from agent-based models, can yield a more accurate model of social phenomena.
This paper used agent-type modeling to analyze the role of information that the government transmits to the public in mitigating COVID. 🇨🇳
Read this book to understand how to use agent-based models to reconstruct and better learn about the human past. 🇩🇰 🇺🇸
This review compiles the different ways in which agent-based modeling (Agent-based Model) has been used in migration and modern slavery research, and help researchers use agent-based models in social science research. 🇬🇧
In agent-based models, individual agents interact with each other and their environment using certain simple rules. These interactions in the face of a changing environment can result in the agents changing their behavior in response to more information. In the realm of computer science machine learning-based inference models generally learn agents’ behavior patterns and can improve decision making. Naturally, there seems to be a connection.
In this paper, the authors provide a comprehensive review of applying machine learning to agent-based modeling. They explain four scenarios based on the level and way in which machine-learning is helping agents make decisions. For each of the four scenarios, the authors specify the related algorithms, frameworks, procedures of implementations, and further multidisciplinary applications.
This is helpful to evaluate our progress in the fields of both agent-based modeling and machine learning, and also determine the steps ahead. The authors mention that currently there is a lack of quality data to train the machine-learning algorithms to be able to help the agent-based models. Addressing these challenges would help lay the foundation for more synergies ahead.
Researchers describe how architectural plans can be generated using a combination of agent-based models and deep learning algorithms. 🇮🇷
This paper explains a new approach that uses agent-based modeling to increase performance in logic programming in a specific type of neural network. 🇲🇾
Systems where there are too many agents interacting with each other over a long period of time can be difficult to model and understand. This paper shows an efficient method of simulating such models. 🇩🇪
(F) BioDynaMo: a modular platform for high-performance agent-based simulation (🇨🇭 🇳🇱 🇨🇾 🇬🇧 🇨🇳)
Biological systems are complex. They consist of multiple interacting parts, and often a bottom-up approach works better for modeling them compared to a top-down one. Agent-based models are exactly that, and they offer the additional advantage of being run entirely on computers, as opposed to running physical experiments.
Researchers have developed a simulation platform called BioDynaMo, which is specifically for creating agent-based models to be of use in biological sciences. The project is open source, and the authors have further tested it on three different test cases: neurology, oncology, and epidemiology. The performance results of BioDynaMo show that it performs up to three times faster than current baselines.
The researchers mention that currently, BioDynaMo is being used to gain insights into retinal development, cryopreservation, multiscale (organ-to-cell) cancer modeling, radiation-induced tissue damage and more. They believe that drug development could also benefit by using this tool instead of in vitro experiments. Agent-based modeling seems to be a promising tool for the future of biological research.
This review summarizes how agent-based models and machine learning have been integrated in biological research, from the cellular scale to population-level scale epidemiology. 🇺🇸
Researchers use agent-based models to model cells as agents and study the response of the immune system during different types of infection. 🇺🇸
This paper advocates the use of Agent-based Models in medical science research studying prevention of adverse health outcomes. 🇺🇸
Special mention
This paper reviews the use of Agent-based Models in the history of social science.
Research Community Map
This month’s newsletter featured research from 6 US States, and 7 countries
Spread the Word!
Thank you so much for reading! Please share this newsletter with anyone who would be interested. Press the button below to share!