ERC SEGUE Team at the eX Modelo Workshop in the Paris, France 2023

SEGUE @ ODISSEI Conference for Social Science in the Netherlands 2023

On November 13th and 14th, 2023, our team member, Bayi Li, participated in a teaching workshop on modelling and simulation. The workshop aimed to empower participants to explore and validate their agent-based models.

The workshop, led by a team of researchers with recognised expertise in interdisciplinary practices, provided step-by-step advanced methods for model exploration over the first two days. Prior to this, theoretical courses were conducted to explain what Agent-Based Modelling (ABM) is, its purpose, and the characteristics of a good model. Special attention was given to introducing the terms “stochasticity” and “heuristics.” Subsequently, the workshop delved into the core topic of automatic model optimisation, elucidating the details of the Non-dominated Sorting Genetic Algorithm.

A key tool highlighted during the workshop is OpenMOLE, a versatile software for both forward and backward methods in model exploration. Throughout the sessions, the forward method, involving sensitivity analysis, and the backward method, focusing on calibrating tunable parameters in the model, were demonstrated. Additionally, the workshop covered scenario modelling with free parameters in OpenMOLE.

One of the notable advantages of OpenMOLE is its ability to serve as a container for models written in different languages, linking the analysis method with the Agent-Based Model (ABM) module(s). This allows for the integration of multiple models in various languages into a single script, promoting reusable code and a modular design approach. Other strengths of OpenMOLE include seamless integration with High-Performance Computing (HPC) and the capability of distributed computing.

The final day of the workshop featured practical sessions where participants applied these methods to various simulation models and programming languages. The OpenMOLE technical team provided support for participants working with their own models.