Since the Covid-19 pandemic, agent-based modelling (ABM) surfaced in many academic circles as it was used it to predict the infection rates and hospitalisations but it's more than a trendy method. ABMs give us a way to explain and express how interactions between the different parts of a system bring about emergent behaviours such as societal change. From explaining segregation, wealth inequalities, opinion polarisation to civil violence, this complexity science approach has a lot to offer in the social sciences. I first learned about ABMs during my Masters as I came across Thomas Schelling's segregation model developed on a checkerboard. There was such elegance and simplicity to that model where a single parameter capturing people's tolerance towards others could explain such a complex phenomenon. I wanted to know more, and I did.
The journey to learn more about agent-based models was a treacherous one. At times, I felt like Odysseus on his journey as I faced different challenges along the way. These ranged from access to expertise, support, high computing power to facing sceptical colleagues and finding a community to belong to. The first thing to keep in mind is that you're going to be spending a lot of time just sitting, thinking about your model, and drawing mind maps about what you're exactly modelling before you get to write any code. This is the most important and fun part of the process so remember to be kind to yourself.
To develop ABMs, it's helpful to take specialised programming courses to design and implement your model. These courses use NetLogo, the most popular and accessible software to learn ABMs, in my opinion. Alternatively, you can use Python's MESA and Agentpy modules or Julia's Agentjl package, but these may require existing programming knowledge and there aren't as many specific courses on these. Moreover, depending on your model size you may need access to a high-performance computing (HPC) cluster to run your simulations. Both, Durham's Advanced Research Computing and Glasgow University, my previous institution, had these facilities which made my life a lot easier. These are just things to consider when following the ABM path but that shouldn't discourage you if you don't have such access or you've never programmed before. You can still learn and run your model on your computer without the world falling apart. It will just take you slightly longer.
Using this approach in my research opened my eyes to the many possibilities it offers while enabling me to go beyond what I originally set out to do in my PhD. Conceiving social science phenomena as complex systems is a fascinating take on how the pieces of the puzzle you're researching fit together. In my case, I was interested in understanding what was it about national identity, in a context where independent movements were present, that made people want to protest. Was it their social networks, the information they were exposed to on social media or their grievances about the issue? So I built an agent-based model to explore how these interconnected components of the system resulted in the emergence of national identity polarisation and protest mobilisation. Using an ABM I was able to provide a possible and plausible explanation for how social media and social networks promote national identity polarisation which in turn promoted protests, not possible with other methods.
Yet, using ABMs in social sciences has always felt like being in no man's land. You don't fully fit in your core discipline, political sciences in my case, but also you don't quite fit in the complexity sciences with the physicists and mathematicians. Navigating that environment during my Masters, but especially my PhD, was challenging in many ways. One of the biggest challenges, besides designing my ABM, was the scepticism in academia to this "new" method. This issue is deeper than me using ABMs for my research but rather evidences the fact that people are very comfortable in their disciplinary silos and often aren't keen to get exposed and listen in to a different perspective. This is problematic, not just for PhD students but also for early-career researchers who are trying to find their voice, as often, all you need is someone that wants to listen to you talk about your research.
You might be wondering, where to from here? While it was discouraging, it pushed me to find other people that used this method or any under the complexity science umbrella. These are soft systems thinking, system dynamics, and network science. The first two, in particular share many of the underlying principles and properties so there's a lot of common ground. My colleague Elizabeth Inyang uses soft-systems methodology applied to public health issues and we got thinking we needed a forum for people like us that might be using these methods or are keen on learning about them. The result of this was an informal and diverse PhD-led systems group where we presented and discussed our own research, evaluated research articles and delivered internal workshops at Glasgow University on complexity science research methods and public health. At Durham, you can find the Durham Research Methods Centre (DRMC) which promotes such interdisciplinary discussions and could be a good starting point to meet other colleagues and students that use or might be interested in using ABMs in their research. Their Methods Café are a great opportunity to informally meet other colleagues with the shared methods interest but different topics besides the traditional seminars during term-time.
Just like Odysseus in the end, minus the decade part of the story, I finally found my home in a supportive, interdisciplinary, and international scholarly community. My advice to any academic, regardless of their career stage, interested in this approach and method is to reach out beyond your department and institution. One point of reference is the European Social Simulation Association (ESSA) which organises annual conferences as well as ABM summer schools. They're an active and welcoming community of scholars with dedicated interest groups. Their hybrid Social Simulation Conference is taking place this year in Glasgow just before their ABM summer school in Aberdeen.
Overall, my journey of learning and implementing agent-based models has been a tough one with many rewarding moments along the way. From the little eureka moment as I was writing the agent rules that govern the simulation to the feedback and encouragement I got at the 2022 Social Simulation Conference in Milan last September or presenting at the Institute for Analytical Sociology (IAS) of Linköping University in February and all the other moments in-between. But most importantly, it has brought me to my current role as a research associate in computational social sciences at Durham University working with Dr. Jen Badham on developing realistic artificial social networks (RASN) for simulation models. All in all, there are many good things about ABMs despite of some of the barriers but relying on your supervisors, colleagues, and friends makes your journey worthwhile.
Research Associate in Computational Social Sciences
Fellow at the Durham Research Methods Centre (DRMC)
cristina.chueca-del-cerro@durham.ac.uk | @ChuecaCristina
Comments
Post a Comment