Alexandre Miguel da Silva Pires

Indirect reciprocity (IR), where reputations indicate with whom to cooperate or defect, is one of the key mechanism for supporting prosocial behavior among unrelated individuals. This mechanism has been studied in the context of human-human interactions. However, as artificial intelligence systems continue to be deployed, the introduction of artificial agents in society has the potential to fundamentally alter reputations' assignment and spread, affecting cooperation dynamics. Artificial agents are fundamentally different from humans, and can vary in their characteristics: they can have a wide social reach, be centralized (e.g., chatbots) or decentralized (e.g., local LLMs), be physical or virtual, and more. Despite this, like humans, artificial agents can also assign and spread reputations, and cooperate or defect. My thesis focuses on creating a framework to study the possible impacts that artificial agents have on human cooperation through IR, using both theoretical models and user studies.