Behave Summer School 2020
Jointly organised by Behave Lab, ESSA-European Social Simulation Association, GECS–Research Group on Experimental and Computational Sociology (University of Brescia), the ESLS PhD Programme in Economic Sociology and Labour Studies (University of Milan) and supported by Fondazione Grazioli and Collegio Universitario Luigi Lucchini di Brescia, this school aims to train students on Agent-Based Models (ABM) in NetLogo by using modelling examples from social science research.
- Lectures + hands-on training with two virtual rooms in parallel for student assistance
- The first course addressed to beginner or experts needing a refresh, the second to advanced training on model calibration and ouput statistical analysis, with an intermediate sunday tutorial on R for beginners.
- Some classic ABMs will be coded from scratch in NetLogo. No previous coding skills required during the first course.
- Assistance and customized counselling on personal research projects during the course with leading experts.
- Personal project presentations.
- Students will be provided with the theoretical background on the use of ABM in social science research and will learn how to develop an ABM from scratch. No prerequisite on computing is needed. Students will be connected via ZOOM and will be trained with their own laptop. Students are also encouraged to develop a customized project starting from a personal research idea: bring your own model or data if you have, and we will help you! During the advanced course, students will be trained on empirical calibration of parameters with quantitative and qualitative data, validation techniques and model documentation. The last slot of the advanced training will include a session on how to survive peer review and editors when trying to publish ABM studies in scholarly journals.
The 2020 faculty includes:
Federico Bianchi (GECS, University of Brescia, Italy)
Ernesto Carrella (University of Oxford, UK)
Simone Gabbriellini (University of Trento, Italy)
Nicolas Payette (University of Oxford, UK)
Gary Polhill (The James Hutton Institute, Aberdeen, Scotland, UK)
Flaminio Squazzoni (University of Milan, Italy, School director)
ESSA-BEHAVE Summer School 2020 | Introductory Course | ||
---|---|---|---|
Monday, 31 August 2020 | Tuesday, 1 September 2020 | Wednesday, 2 September 2020 | |
9:30AM-10:AM | Welcome | ||
10AM-11:30AM | The Theory and Methodology of Agent-Based Modelling (Flaminio Squazzoni, University of Milan). This lecture provides an introduction to the application of agent-based modelling in the social sciences. The focus is on: explanatory and counterfactual analysis; implementing and testing hypothetical micro-macro social mechanisms; developing evolution scenarios by parameter manipulation. | Axelrod’s (1997) model on the dissemination of culture in NetLogo (Simone Gabbriellini, University of Trento). This lecture uses a canonical model to train students on agent-based modelling in NetLogo | Axelrod’s (1997) model on the dissemination of culture in NetLogo (Simone Gabbriellini, University of Trento). This lecture uses a canonical model to train students on agent-based modelling in NetLogo |
12-1:30PM | From theory to coding (Simone Gabbriellini, University of Trento). This lecture introduces students to ABM thinking and modelling philosophy. The focus is on how to move from a variable-based to an agent-based modelling approach | Axelrod’s (1997) model on the dissemination of culture in NetLogo (Simone Gabbriellini, University of Trento). This lecture further develops previous steps and advances on the full version of the model | Axelrod’s (1997) model on the dissemination of culture in NetLogo (Simone Gabbriellini, University of Trento). This lecture further develops previous steps and advances on the full version of the model |
3-4:30PM | Introduction to NetLogo (Nicolas Payette, University of Oxford). This lecture introduces students to NetLogo | NetLogo Focus (Nicolas Payette, University of Oxford). This lecture focuses on NetLogo functions and coding tricks learned in the morning | NetLogo BehaviorSpace (Nicolas Payette, University of Oxford). This lecture trains students on how to run simulations via BehaviorSpace (built-in NetLogo function). It uses Alxerod’s model developed in previous lectures and help students to understand how to analyse simulation results. |
4:30-6PM | Introduction to NetLogo II (Nicolas Payette, University of Oxford). This lecture provides a hands-on introduction to NetLogo by presenting basic programming logic and functions. It will help students coding their first model on social collective dynamics. | Counselling. The faculty will help students to define and develop their model by concentrating either on ABM research questions/simulation design strategies or specific modelling details | How to excel doing ABM research (Flaminio Squazzoni, University of Milan). This lecture provides tips and tricks from the viewpoint of the editor of JASSS-The Journal of Artificial Societies and Social Simulation, the flagship journal in the ABM field. |
6-7PM | Counselling. The faculty will help students to define and develop their model by concentrating either on ABM research questions/simulation design strategies or specific modelling details. |
ESSA-BEHAVE Summer School 2020 | Advanced Course | ||
---|---|---|---|
Monday, 7 September 2020 | Tuesday, 8 September 2020 | Wednesday, 9 September 2020 | |
9:30AM-10AM | Welcome | ||
10AM-11:30AM | What are we doing? Some theoretical perspectives on calibration and validation of agent-based models (Gary Polhill, The James Hutton Institute, Aberdeen, President of the European Social Simulation Association). Traditional quantitative modelling focuses on fitting mathematical functions to empirical data. The parameters of these functions adjust the shape the function forms as it maps its domain to its range. All of the mathematics around calibration and validation is thus necessarily focused on finding ‘the right’ values for these parameters and establishing the level of confidence we can reasonably have in the model’s predictions using those parameters. Although these matters are still relevant to agent-based models, there are important differences both in the style of modelling and the contexts in which the models are applied that merit attention, and mean there is more to the story of calibration and validation of ABMs than which function in R you should use. | Estimation of agent-baseds models (Ernesto Carrella, University of Oxford). In this lecture, I will briefly summarise the advances in likelihood-free methods for estimating agent-based models and how they can be used to match parameters to data. Almost all techniques involve tracking a set of summary statistics and trying to minimize the distance between simulated summary statistics and the real ones. I will start by showing the sufficient conditions by which this process “works” (that is, estimated parameters converge to the “real” ones). I will then show a set of techniques to perform this estimation trying to tease out their strengths and weaknesses. I will show however how in practice identification problems are ubiquitous and how one can spot at least the most glaring examples. We will conclude discussing how estimation fits into the process of building simulations and in particular its relationship to validation and theory development more broadly. | Introduction to R and tidyverse for output analysis (Nicolas Payette, University of Oxford). This lecture introduces basic functions of the tidyverse packages for data analysis in R used for simulation output analysis. |
11:30-1:30PM | The challenge exercise (Simone Gabbriellini, University of Trento & Nicolas Payette, University of Oxford). This lecture uses a model to present a puzzle and a challenge to participants. This includes: the design of an ABM to explain an empirical puzzle on different innovation diffusion rates in two empirically-observed social networks. Students will code macro-level initial conditions of the model in NetLogo and will calibrate the model to empirical network data. Students will be familiarised with a git repository for version control of the challenge model | ABM challenge – Mechanism modelling (Simone Gabbriellini, University of Trento). This lecture will help students designing hypothetical mechanisms to explain the challenge puzzle and implement them into the model | ABM challenge – Validation (Simone Gabbriellini, University of Trento). This lecture will present the validation of the model via comparison of simulation results against empirically-observed aggregated data. |
3-4:30PM | Unit testing (Nicolas Payette, University of Oxford). This lecture helps students understand how to test model code | ABM challenge – Parameter sweeping (Nicolas Payette, University of Oxford). This lecture helps students to understand how running simulations of the coded model through NetLogo’s BehaviorSpace and terminal. | Student model/project presentations |
5-6PM | Counselling. This slot provides students the opportunity to benefit from collective/personalised counselling on ABM-related issues, work-in-progress research projects, coding issues. | Counselling. This slot provides students the opportunity to benefit from collective/ personalised counselling on ABM-related issues, work-in-progress research projects, coding issues. | Student model/project presentations |