Boltzmann-type and mean-field models in socio-economic applications

Lead Research Organisation: University of Warwick
Department Name: Mathematics

Abstract

Our interactions and behavioral patterns not only impact our personal environment, but also effect the progress and development of our society. Understanding how the exchange of information and knowledge between individuals impacts the productivity of an economy and which incentives promote life long learning and prolonged productivity is of great importance and has stimulated research in the field of social sciences, economics and more recently applied mathematics.

A main challenge corresponds to the problem of deriving mathematical models for macroscopic quantities, for example the productivity of a society, from microscopic interactions. A related problem was studied by Ludwig Boltzmann, who wanted to describe macroscopic quantities of a gas, such as the pressure or the temperature, based on the microscopic collisions of gas molecules. He proposed that the macroscopic behavior can be deduced by studying the average behavior of the system, laying the foundations for kinetic theory. Similar ideas can be adapted to understand how individual interactions affect the overall behavior of a large group. But human interactions are rather complex, often based on personal preferences and/or rational decisions. Rational decision making corresponds to considering pros and cons, while trying to find the optimal strategy. This problem is well known in mean-field game theory, which studies strategic decision making in large groups of interacting individuals.

In this project we aim to develop mathematical tools to describe and analyse trading mechanisms and their impact on the dynamics of the price, using ideas proposed in the field of kinetic theory. A second goal corresponds to the coupling of kinetic and mean-field game theory to understand how strategic decisions influence the overall behavior of large groups. This approach will be used for example to analyse the effect of knowledge exchange and dissemination on the overall productivity of an economy. The developed techniques and results allow us to identify driving factors for sustained economic growth. As a part of this project we will develop numerical algorithms, which shall support the theoretical results but also give implications about the dynamic behavior in more complex situations.

The research will be carried out at the Mathematics Institute at the University of Warwick by the PI working with collaborators from prestigious overseas and national universities. These experts in the field of nonlinear partial differential equation theory, mathematical finance as well as economics will be actively involved in the project, and will help to develop and disseminate the obtained results.

Planned Impact

Understanding how complex economical and behavioral interactions influence the dynamics of a price or economic growth are questions which have challenged researchers in the fields of economics, social sciences and more recently applied mathematics. The main goals of this research project address these challenges, by developing a mathematical framework which shall give novel insights into the driving forces of these complex processes. The proposed project outcome will benefit the following groups:

*) Researchers in the field of kinetic and nonlinear PDE theory
*) Researchers from financial mathematics studying price dynamics or the distribution of wealth
*) Economists working on multi-agent systems
*) Researchers in social sciences analysing opinion dynamics or consumer behavior
*) On the longer time-scale opinion polling and development aid policy

The impact on researcher in kinetic and nonlinear PDE theory is dercibed in the Section 'Academic beneficiaries'. Boltzmann-type models have been used successfully to understand the distribution of wealth in a society or the dynamics of a price traded in an economic market. These models can be understood as building blocks to address more complex problems and provide versatile mathematical tools which can be used to analyse similar problems in socio-economic applications. The proposed coupling to mean-field game models will impact research on the dynamics of multi-agent systems in economics and social sciences. It provides the necessary mathematical tools for the development of quantitative models,which allow to identify the main driving forces of complex processes. Our collaborators are among the leading experts in the field, which will help to disseminate our work to the target audience and promote these models in their respective communities.

The results of the project can be put into use to understand complex dynamics in our society in the future. For example by analysing the impact of personal communication and information transfer on opinion dynamics or consumer choices.

Publications

10 25 50
 
Description We have worked on different kinetic and mean-field models for interacting particle systems, which were motivated by real life applications. In a first project we studied the alignment behavior of pedestrians, which try to reach a certain target but at the same time avoid collisions with others. We were able to prove that groups with different objectives, that is for example different targets or walking directions, segregate and directional bands form. These findings were also confirmed by numerical experiments. A second project focuses on the analysis of a mean-field model in which particles diffuse through a porous medium. Here we are interested in the existence of solutions and the long time behavior - first results were submitted for publication and numerical simulations confirmed the expected segregation effects. We are currently working on mean-field models for interacting particles systems, where we observe segregation effects and sorting behavior.
Exploitation Route The proposed models can serve as a starting point for the development of more realistic descriptions - however the developed analytic and numerical tools give insights into the dynamics and identify what interactions are necessary to observe segregation in pedestrian flows or in a society. They can also contribute to understand the long time behavior of large interacting systems - a question, which is difficult to answer when studying the dynamics on the individual level.
Sectors Communities and Social Services/Policy

Digital/Communication/Information Technologies (including Software)

Transport

 
Description Kinetic opinion formation models for digital societies
Amount £12,000 (GBP)
Organisation The Royal Society 
Sector Charity/Non Profit
Country United Kingdom
Start 04/2022 
End 04/2024
 
Description Collaboration with Andrew Stuart (Caltech) 
Organisation California Institute of Technology
Country United States 
Sector Academic/University 
PI Contribution Connecting mean-field models for collective dynamics with micrscopic data using the Bayesian framework.
Collaborator Contribution A. Stuart is one of the leading experts in Bayesian inverse problems.
Impact *) Royal Society International Exchange Grant *) Two papers
Start Year 2017
 
Description Collaboration with Bertram Duering (University of Sussex) and Marco Torregrossa (Pavia) 
Organisation University of Sussex
Country United Kingdom 
Sector Academic/University 
PI Contribution We started to work on a kinetic model to rank player. It is based on the so-called ELO ranking system, which is used the International Chess Foundation, the National Football League or the FIFA. Players are characterised by their intrinsic strength and ranking - both of them change to interactions. The proposed model allows us to study the long-time dynamics of individual rankings depending on the learning and training mechanisms.
Collaborator Contribution The partners supported me in the model development and the analysis. Their contributions allowed to make quicker progress and helped with the development of numerical tools.
Impact We are currently working on a paper, which should be submitted within the next month.
Start Year 2017
 
Description Collaboration with Matt Barker and Pierre Degond on mean field game models 
Organisation Imperial College London
Country United Kingdom 
Sector Academic/University 
PI Contribution Investigate the difference between mean-field games and the best reply strategy
Collaborator Contribution Analytical aspects on best-reply strategy and similar approximation techniques for MFG.
Impact Submitted a paper investigating stationary solutions to MFG and BRS models.
Start Year 2018