PhD Seminar • Artificial Intelligence • Information Transmission on Social Networks with Preference Homophily

Tuesday, April 9, 2024 9:00 am - 10:00 am EDT (GMT -04:00)

Please note: This PhD seminar will take place online.

Valerie Platsko, PhD candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Kate Larson

Social networks, from online social networks to institutional structures governing who communicates directly, shape communication and play a significant role in the ability to communicate true information in the presence of competing incentives. Rational agents, even if they may not spread false information, may have incentives to withhold information or to provide imprecise information if a collective decision will be made based on the shared information and informed agents have heterogeneous preferences. These strategic choices to withhold information are undesirable in many contexts and result in less-informed decisions, even though they do not result in active misinformation. This work considers the factors that influence communication, such as the distribution of agent preferences in the network and knowledge about neighbours in the network,  when agents are self-interested but short-sighted instead of fully-rational.

A main result is that limited visibility in the network, for example of neighbours’ preferences, increases the ability of self-interested agents to communicate under some network structures but decreases it under others. This occurs based on differences between the local neighbourhoods of agents with preferences that are a minority in the whole population. This provides potential insights for the design of social network structures that will enable more communication to occur.