Modeling the Diffusion of Information in Multi-Agent Societies: Evaluating Rumor Models with a Group-Based Simulation
Liljenmaa, Jesse (2023)
2023
Tietojenkäsittelyopin maisteriohjelma - Master's Programme in Computer Science
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2023-06-13
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202306076595
https://urn.fi/URN:NBN:fi:tuni-202306076595
Tiivistelmä
Rumor modeling is the mathematical study of rumors and their diffusion in populations for the purposes of understanding the underlying phenomena, such as how far away and how fast a rumor spreads, and the incidental posterior effects, such as shifts in social, economical, or political preferences. Rumors are traditionally studied with compartmental models that have been developed based on previous epidemiological models, which has influenced the way researchers see rumors as idea-based diseases.
Rumor diffusion is ordinally simulated with the help of social networks, where a rumor propagates through a graph of vertices and edges representing social connections between entities such as humans. These social networks can be constructed either with algorithms or based on real-life data. However, both approaches have weaknesses that distance them from practical application. For the former, theoretical, social networks, there is considerable difficulty in their usage in real-life situations. However, for the latter, practical, networks, this is not a problem, but the results are not able to be abstracted to apply in the general case.
The goal of this thesis is to explore the middle ground between theoretical and practical social networks. With this in mind, this thesis proposes a novel way to evaluate rumor models through a group-based simulation. The contributions of this simulation to the field of rumor modeling are two-fold: first, three classical rumor models are brought to the modern, social media age by modifying them to function with groups, i.e., many-to-many relations instead of pairwise connections of the old. Next, a new parameter is introduced for rumor modeling in the form of the environment size, which affects the density of the agents.
To confirm the practicality of the simulation as well as the performance of the rumor models considered, each model is evaluated with a multitude of test cases in order to deduce the full nature of rumor diffusion within this environment. The results indicate that the updated models work well, and that each scenario contains unique features, proving good opportunities in their analysis. Additionally, the added dimensionality of the parameters provides more accurate means to model and study rumors in depth.
Rumor diffusion is ordinally simulated with the help of social networks, where a rumor propagates through a graph of vertices and edges representing social connections between entities such as humans. These social networks can be constructed either with algorithms or based on real-life data. However, both approaches have weaknesses that distance them from practical application. For the former, theoretical, social networks, there is considerable difficulty in their usage in real-life situations. However, for the latter, practical, networks, this is not a problem, but the results are not able to be abstracted to apply in the general case.
The goal of this thesis is to explore the middle ground between theoretical and practical social networks. With this in mind, this thesis proposes a novel way to evaluate rumor models through a group-based simulation. The contributions of this simulation to the field of rumor modeling are two-fold: first, three classical rumor models are brought to the modern, social media age by modifying them to function with groups, i.e., many-to-many relations instead of pairwise connections of the old. Next, a new parameter is introduced for rumor modeling in the form of the environment size, which affects the density of the agents.
To confirm the practicality of the simulation as well as the performance of the rumor models considered, each model is evaluated with a multitude of test cases in order to deduce the full nature of rumor diffusion within this environment. The results indicate that the updated models work well, and that each scenario contains unique features, proving good opportunities in their analysis. Additionally, the added dimensionality of the parameters provides more accurate means to model and study rumors in depth.