Analysis of agent-based opinion formation models

Carlos Andrés Devia
Ph.D. Candidate, Delft University of Technology
Netherlands
DEIB - Seminar Room "N. Schiavoni" (Bldg. 20)
November 8th, 2022
10.30 am
Contacts:
Fredy Orlando Ruiz Palacios
Research Line:
Control systems
Ph.D. Candidate, Delft University of Technology
Netherlands
DEIB - Seminar Room "N. Schiavoni" (Bldg. 20)
November 8th, 2022
10.30 am
Contacts:
Fredy Orlando Ruiz Palacios
Research Line:
Control systems
Sommario
On November 8th, 2022 at 10.30 am Carlos Andrés Devia, Ph.D. Candidate at Delft University of Technology of the Netherlands, will hold a seminar on "Analysis of agent-based opinion formation models" in DEIB - Seminar Room.
Agent-based opinion formation models embed the psychological traits of individuals to explain and mimic the mechanisms that produce opinion evolution in society. These models are able to produce a wide range of outcomes, and the characterization of the possible qualitative behaviors that they can generate is an open problem. Which types of opinion distributions can a model produce starting from a given initial condition? Can an initial opinion distribution be transformed by the model evolution into a significantly different configuration? How do these outcomes depend on the agent parameters?
In this presentation, a novel methodology is proposed, which aims to solve this problem by identifying different qualitative opinion distribution categories (Perfect Consensus, Consensus, Polarization, Clustering, Dissensus) and investigating the possible transitions between different categories, also exploiting suitable representations that concisely capture the general level of agreement in the population. Simulation results showing the application of the proposed methodology to two classical and two new opinion formation models will be discussed, together with potential adaptation to agent-based models in other research fields.
Agent-based opinion formation models embed the psychological traits of individuals to explain and mimic the mechanisms that produce opinion evolution in society. These models are able to produce a wide range of outcomes, and the characterization of the possible qualitative behaviors that they can generate is an open problem. Which types of opinion distributions can a model produce starting from a given initial condition? Can an initial opinion distribution be transformed by the model evolution into a significantly different configuration? How do these outcomes depend on the agent parameters?
In this presentation, a novel methodology is proposed, which aims to solve this problem by identifying different qualitative opinion distribution categories (Perfect Consensus, Consensus, Polarization, Clustering, Dissensus) and investigating the possible transitions between different categories, also exploiting suitable representations that concisely capture the general level of agreement in the population. Simulation results showing the application of the proposed methodology to two classical and two new opinion formation models will be discussed, together with potential adaptation to agent-based models in other research fields.
Biografia
Carlos Andrés Devia received a M.Sc. in Electrical Engineering from Pontifical Universidad Javeriana, Colombia where he worked on optimal control and model predictive control in unmanned aerial vehicles.
Since 2019 he is a Ph.D. Candidate at Delft University of Technology, the Netherlands.
His Ph.D. is on the study of interconnected dynamical systems from a theoretical and numerical perspective. His current research focuses on the analysis and development of agent-based opinion formation models.