Modeling Opinion Diffusion over a Social Network using a Genetic Algorithm
Kara Johnson, Nicole Carnegie, Ralph BarnesThe DeGroot model for opinion diffusion over social networks dates back to the 1970s and models the mechanism by which information or disinformation spreads through a network, changing the opinions of the agents. There exists extensive research about the behavior of the DeGroot models and its variations over theoretical social networks with specified characteristics; however, research on how to fit these models using data collected from an observed network diffusion process is much more limited. Because of these current limitations, DeGroot models and their variants are an untapped resource for a variety of applications including mitigating the spread of fake news, assessing the efficacy of advertisements, and determining the most efficient allocation of resources in political campaigns.
While a variety of other methods can be used to assess more holistic features of opinion diffusion over a network such as how quickly opinions change and spread through a network, the DeGroot model allows for the identification of influential or high-value agents within the network. Agents of particular interest include those who are a catalyst for change due to their opinions being valued by other agents or agents who are particularly open or resistant to changing opinions.
Existing methods require datasets that are prohibitively large for many social science applications. This talk will present a novel genetic algorithm capable of fitting opinion diffusion models with small datasets and missing observations, including those where there are more parameters than data points. The algorithm will be demonstrated on data from a study investigating the process by which minority beliefs become the belief of the majority on an artificially-created social network.