This blog describes our reflections of the Social Simulation Week 2020 (14-18 Oct 2020), an event of the European Social Simulation Association (ESSA) organised by the Behave Lab by Nanda Wijermans and Udita Sanga. Nanda has been associated with ESSA for a long time (almost 12 years!) while Udita attended the annual conference for the first time as an introduction to the wider ESSA community.
Nanda: While getting back to work the monday after the Social Simulation week my brain is trying to catch up with what happened. It was exciting & exhausting, stimulating & mind-numbing, a weird mix. The annual conference was scheduled to be organised in Milan this year but due to COVID 19, the conference was moved to an online platform. It was a big self-organised fest, where each session was fully in the control of each session proposer. Leading to some sessions being like a conference track with a line up of talks spanning over the whole week (ABM & Networks) to a 3hr slot on a day with discussion groups in breakout rooms. Quite the selection for a proper science-binge (see programme).
Udita: A virtual conference was perhaps a blessing in disguise. It made me wonder if virtual large-scale conferences will become more common now; given their higher flexibility, accessibility, and low carbon footprint than standard academic conferences. I did miss the coffee breaks though and the opportunity to exchange ideas, impressions, and make connections with other participants. Yet, the virtual conference was a peek into a different way of interacting with each other and the community, especially the chat sections, use of online polling, and breakout groups for shorter discussions. However, the 3-hour sessions were a bit too long and as Nanda mentioned, did tend to get exhausting by the end of the day. The online multi-session formats would have benefitted from reducing the workshop session durations and adding more short breaks to allow participants to manage zoom fatigue.
Overall, many sessions had a great appeal, there were opportunities to support and share work in progress (ESSA@work), support & tools for social simulation development (Creation of NetLogo ABM from a Formalised ODD, Julia, A Software Architecture for Multi-theory Mechanism-Based Social Systems Modelling in ABM practices for model sharing), but also methodological reflections (Integrating Qual and Qual evidence, Boundaries of ABM). There were some usual suspects (ABM & Networks, Simulations in economics) and some new themes (changes in consumption habits and production, simulation in times of COVID-19, health behaviour modelling, ABM for prediction; polarisation, othering and social identity, ABM and games, Model discovery)
Some talks we liked:
Keynote Maja Schlüter: “Theorizing about social-ecological phenomena through collaborative modelling”(video): Apart from obviously loving her work while being part of it, we particularly liked her talk because it highlighted some reflections that were just meaningful to share and was inspiring to listen to. For example, she coined and distinguishing the work we do as science as teamwork rather than individual work; she gave an overview of the different types of work done and our rather different take on using empirics in models; Maja also highlighted the need for better ‘intertwinedness” of social and ecological systems in SES models and for generating novel questions that help understand causal mechanisms and processes within a SES.
Keynote Giulia Andrighetto: “Understanding human cooperation through natural and artificial data” (video): she talked about her really cool 30(!) day lab experiment on norm emergence and stabilisation. This is one of the few studies I know of that also combines behavioural experiments with ABM. Quite exciting and also just a reminder for ourselves how amazingly cool experiments our collaborators Therese Lindahl and Caroline Schill do by ‘making things complex’ in ways that we need (e.g. communication, resource dynamics, etc) – (see also AgentEx project).
Talk by Jens Madsen: “Cognitive psychology and behavioral sciences for the modeling of individual behavior” focused on the importance of being human in ABMS. This talk resonated with the work we do in fisheries (e.g. FIBE, OctoPiNTS, CrossFish, FishMob), so that is just stimulating and makes us want to get in touch. They do this within the scope of the Poseidon model and focus on reflection, a more realistic representation of individual and collective fishing effort. He told about the case study work done in Indonesia, by talking to fishers learning about their behavioural assumptions, and going beyond just the economics. Exactly the kind of work we are advocating for (Wijermans et al 2020). We cannot wait for that paper to be out!
Talk by Sukaina Bharwani: “Adding value to social simulation models using qualitative evidence”: because of the way she uses a participatory and collaborative process for formalising qualitative data using agent-based modelling, empirics which is more applied towards supporting local decision making. Very exciting and many links to the way we work and definitely a good person to connect to.
Some sessions we liked
Polarisation, othering, and social identity (Nanda)
More and more attention within the community is going to be able to explain polarisation and group dynamics in general. What does it mean to be part of a group and how does that affect behaviour? It is in line with the wish to model behaviour in context (link blog SSC2018). Since this session was a collaboration of two recent subgroups in the ESSA community one on Polarisation and Othering and Social Identity Approaches for ABM (SIAM) it gave space to introduce these concepts, see some modelling work going on it is but also simply the presence of these networks that are open for anyone to join. The discussions were lively and showed that the interest and relevance of such models are carried widely.
Games and Agent-Based Modelling- Investigating Synergies (Udita)
I was happy to see a separate session on games and ABM and that the niche synergetic space between games and ABM is gathering wider interest of the ABM community. Works by Christophe LePage have been a source of inspiration in my own research and it was wonderful to participate in the session. The discussions on the use of games as research tools, challenges, and benefits of game design and ABM development offered a lot of food for thought; especially as we engage in discussions on reflecting more on improving transparency and clarity on model structure concepts and model discovery
Integrating Qualitative and Quantitative Evidence using Social Simulation [video] (Udita & Nanda)
I think the SIMSOC community widely acknowledges the advantages and challenges in integrating qualitative /quantitative data in the simulation of social phenomena; and the session reflected this. My (Udita) main takeaway from this session was that this also is an opportunity for a deeper reflection on research paradigms and philosophies behind qual and quantitative sciences (e.g. objective vs subjective interpretation; generalizability, predictability ) and what that means for the models we build. This would include establishing better practices for the effective integration of quantitative and qualitative data within the ABM community.
From my side (Nanda), being part of this discussion for longer there is an increasing need for more deep discussion about what is really the issue we face and want to tackle. The different takes and definitions of Qual vs Qual, how Qual is always part of Quan are for my -want to do something with it- mind not enough. What do we mean with integrating?, how do different projects approach this? how can we enable talking to each other when integrating my way is not your way and what can I learn from you by being more explicit? I am eager for more and will ask for more next meeting, but as Udita said, such an important topic and so happy that Melania Borit with Bruce Edmons continuously create this space for the community. (for more info on this group: see ESSA Special interest group Qual2Rule)
Agent-based modelling can be used for prediction in complex social systems (Udita)
The format of the session was in the form of a debate where speakers argued for or against the motion. It was one of the most interactive and passionate sessions I’ve attended; and it seems like the Social Simulation community is quite polarized about what ABMs are useful for ( predictive modelling vs exploratory modelling) (PS: a poll was conducted among the 90 or so participants and 50% thought prediction was possible; 30% voted No and 20% were undecided) The debate mainly revolved on “IF” prediction of complex systems is possible; there were discussions on the need and scope for a more nuanced discussion of the types of models or contexts where prediction may be possible (or even useful) and those where it might not be. An interesting future topic of discussion proposed at the session was “Limits of prediction” and I look forward to this session whenever it happens!
The invited talks are available online, including the Welcome Talk by Gary Polhill; “Theorizing about social-ecological phenomena through collaborative modelling” by Maja Schlueter, “Agent-Based Modeling: Backward and Forward” by Josh Epstein, “How to create generalizable AI?” by Anima Anandkumar , “Understanding human cooperation through natural and artificial data” by Giulia Andrighetto and “What Social Simulation can tell Artificial Intelligence – and vice versa” by Petra Ahrweiler.
We could talk/type so much more, there were sooo many nice talks and sessions, this is just the tip of the iceberg that we just highlighted to give an impression. Overall takeaway: it was a great week for learning and engaging with relevant topics in Social Simulation and are very keen on learning from this week and organise another online simulation week in spring (Nanda is involved ?).