Please enable JavaScript for this site to run correctly.

Bayesian Agent-Based Population Studies

Posted 2021-06-03 13:52:10 by Admin

We are delighted to announce that the course on Agent-based Modelling for Social Research, originally envisaged as a short summer course, was successfully delivered in the online format in November 2020.

The main aims of this course were to familiarise the participants with the most recent advances in building, analysing and documenting agent-based models of social processes. During the course, we covered aspects related to the choice of modelling language and environment, tailoring models for specific research purposes, statistical analysis of model results and key principles of experimental design, inclusion of realistic cognitive assumptions in models, and documenting the modelling endeavours by using a variety of approaches. The course was aimed particularly at PhD level students and early career researchers, with some prior experience with coding and interest in computational modelling in social science. The programme details are listed below.


Preliminary programme (all times GMT)

Day 1 (Tue, 3 November) – Live sessions

13:00 – 14:00              Keynote lecture (Professor Alexia Fürnkranz-Prskawetz, Vienna University of Technology), open
14:10 – 14:20              Opening and welcome. Course preliminaries and preparation
14:20 – 15:00              Informal icebreaker session with course participants only

Day 2 (Wed, 4 November) – Live sessions, with pre-recorded background videos and reading

13:00 – 13:25              Introduction to agent-based modelling
13:25 – 13:50              Introduction to Julia and ML3
14:00 – 15:00              Small groups 1: Setting up. Advisory meetings on selected topics

Day 3 (Wed, 11 November) – Live sessions, with pre-recorded background videos and reading

13:00 – 13:25              Statistical aspects of agent-based modelling (Jason Hilton, Jakub Bijak)
13:25 – 15:00              Small groups 2: Model development, implementation and analysis

Day 4 (Wed, 18 November) – Live sessions, with pre-recorded background videos and reading

13:00 – 13:25              Towards psychological realism in agent-based models (Toby Prike)
13:25 – 13:50              Documentation: ODD and provenance (André Grow, Oliver Reinhardt)
14:00 – 15:00              Small groups 3: Cognitive elements and model documentation

Day 5 (Wed, 25 November) – Live sessions

13:00 – 14:00              Group reports (Plenary: all groups)
14:10 – 14:40              Stocktaking and reflection (Jakub Bijak)
14:40 – 15:00              Virtual coffee and farewell

We expect that the participants commit six hours per week (24 hours in total), with two scheduled contact hours, one more unscheduled contact hour for milestone meetings in small groups, in weeks 2, 3 and 4, and three hours of work on the mini-projects, individually and in small groups.

The deadline for applications passed on 29 February 2020, and successful applicants were already notified about acceptance. We regret that due to high demand we are unable to offer additional places on this course.

Joining instructions will be sent to registered participants in mid-October 2020, together with links to pre-recorded lecture videos and practical information regarding the software.

Additional information for the participants:

The hands-on part of the course will include building an agent-based model of selected social processes in small groups, guided by the course tutors. Below we enclose an initial list of topics which we may consider, and welcome other ideas from the participants.

Topic 1. Who marries whom, and why?

Many parts of the world have recently experienced changes in the educational attainment of women relative to that of men, with women surpassing men in terms of participation and success in higher education since around 1990s. One consequence of this reversal is that long-standing patterns of educational assortative mating have changed. In most couples, the wife and husband have similar levels of educational attainment (homogamy). But, in the past, if there was a difference in educational attainment, the wife tended to be less educated than the husband (hypergamy), while today, if there is a difference, it is the other way around (hypogamy). Common explanations focus on the erosion of traditional gender role norms and changes in structural opportunities to marry educated partners, especially for women. In this project, we will develop a marriage market model that makes it possible to study the interplay between gender norms and the structure of the market in affecting marriage patterns. Inspiration and coding examples can be obtained from existing models.

Topic 2. Network effects on migration decisions

Empirical studies show that the presence of social contacts in potential destination countries plays an important role in migration decisions. Furthermore, we know that depending on cultural background and host country policies, arrived migrants tend to exaggerate or downplay their own success when communicating with friends or relatives in their country of origin. This might lead to a situation where the network effect differs along two axes, destination *and* origin, to the point that it might even become negative for some combinations. In this project, we will build a simple model to explore this further.

Topic 3. Agent-based model of political fake news

The influence of fake news and misinformation on politics have increasingly come into focus in recent years. There is empirical evidence that people differ in the extent to which they believe in and/or are influenced by misinformation as well as how willing they are to share misinformation with others. Partisanship is also likely to play a role, with people more willing to share misinformation that is consistent with their political beliefs and/or benefits their political side. In addition to those who are tricked or misled into believing and spreading misinformation, there are also those with ulterior motives who actively seek to create and encourage the spread of misinformation with full knowledge that it is inaccurate. We will build an agent-based model to examine the spread of misinformation.

Topic 4. Can face coverings help limit the spread of infectious diseases?

There are currently ongoing debates in many countries regarding the effectiveness of using face coverings to limit the spread of COVID-19. At the moment, there is still a lot of uncertainty both regarding the spread of the disease and the efficiency of various counter-measures, not to mention the paucity and lack of international comparability of the relevant data, for example due to differences in the testing regimes. In this exercise, we will build a proof-of-concept model trying to demonstrate a possible way of assessing the effectiveness of face coverings by using an agent-based approach. In this exercise, we will look into the properties of social networks, the impact of attitudes and behaviour towards risk, and will conduct a global sensitivity analysis to identify which aspects of the model (parameters) are actually important for the model behaviour and its various outcomes.