Simulating social dynamics: a computer aided analysis of emergent behaviour in multi-agent systems
File(s)
Author(s)
Carver, Alexander Nicholas
Type
Thesis or dissertation
Abstract
There is an existing gap in the literature and research in computational social dynamics in examining
the role of the motivations and aims of agents in existing social models, in particular in regards to how
available information, and the amount of information, is utilised by different agents to achieve their
aims. Instead, agent behaviour is largely operated via simplistic rules imposed by the global model
that do not take the agent’s distinct motivations or available information into account. My research
has sought to counteract this trend in the existing research. I have done this by critically examining the
assumptions of commonly used models such as the Schelling Segregation model, or concepts such
as strategyproofness in peer review systems, and demonstrating that assumptions made about such
behaviour or the utility of such concepts are not as solid as they may appear, showing that the Schelling
segregation model does not necessarily impose global segregation, and that strategyproofness may be
an unnecessarily strong condition when the incomplete information and motivations of manipulating
agents is taken into account. Based on these results, I have combined these into an approach of
examining the role of information and agent motivations in segregation models, which I feel has
not been subject to sufficient scrutiny in existing segregation models. Therefore I have proposed
two new models of segregation, the first of which examines how the varied motivations of agents
may be more complex than simply desiring homophilous groupings and seeks to replicate intra-type
segregation based on a ”segregation by association”. The second segregation model I propose then
studies the role of incomplete information in segregation, and in basing itself on the result from study
on incomplete information in peer review, argues that in a similar manner, we must examine how
agents take decisions with incomplete information, and that socially sub-optimal segregation could be
explained by biases these agents operate upon in the absence of complete information. Overall, I hope
that my research demonstrates the need for social dynamicists to critically question the assumptions
we make and put into models regarding how the behaviour of agents relate to their motivations of
agents and their available information, and if we can design models and theories that explain social
behaviour that take these questions into account.
the role of the motivations and aims of agents in existing social models, in particular in regards to how
available information, and the amount of information, is utilised by different agents to achieve their
aims. Instead, agent behaviour is largely operated via simplistic rules imposed by the global model
that do not take the agent’s distinct motivations or available information into account. My research
has sought to counteract this trend in the existing research. I have done this by critically examining the
assumptions of commonly used models such as the Schelling Segregation model, or concepts such
as strategyproofness in peer review systems, and demonstrating that assumptions made about such
behaviour or the utility of such concepts are not as solid as they may appear, showing that the Schelling
segregation model does not necessarily impose global segregation, and that strategyproofness may be
an unnecessarily strong condition when the incomplete information and motivations of manipulating
agents is taken into account. Based on these results, I have combined these into an approach of
examining the role of information and agent motivations in segregation models, which I feel has
not been subject to sufficient scrutiny in existing segregation models. Therefore I have proposed
two new models of segregation, the first of which examines how the varied motivations of agents
may be more complex than simply desiring homophilous groupings and seeks to replicate intra-type
segregation based on a ”segregation by association”. The second segregation model I propose then
studies the role of incomplete information in segregation, and in basing itself on the result from study
on incomplete information in peer review, argues that in a similar manner, we must examine how
agents take decisions with incomplete information, and that socially sub-optimal segregation could be
explained by biases these agents operate upon in the absence of complete information. Overall, I hope
that my research demonstrates the need for social dynamicists to critically question the assumptions
we make and put into models regarding how the behaviour of agents relate to their motivations of
agents and their available information, and if we can design models and theories that explain social
behaviour that take these questions into account.
Version
Open Access
Date Issued
2022-04
Date Awarded
2023-02
Copyright Statement
Creative Commons Attribution NonCommercial Licence
Advisor
Turrini, Paolo
Publisher Department
Computing
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)