Grand Challenge — Social Networks in the Age of Humans and AI

Social networks are central to how trust, communication, and information flow develop, and they are increasingly shaped by the presence of both humans and artificial social agents. Advances such as virtual reality and large-scale online experiments now make it possible to investigate these dynamics with both experimental control and ecological realism.

Talk 1 — Humans learn to prefer trustworthy AI over human partners (Yaomin Jiang)

Partner selection — choosing whom to interact with — is a building block of social networks and a key driver of cooperation, and often hinges on communication. As artificial agents become more intelligent, autonomous, and persuasive, they compete with humans for partnerships. I present experimental work on a communication-based partner-selection game in simulated mini-societies of humans (N = 975) and AI agents. When AI candidates do not disclose their nature, their good behaviour is often misattributed to humans and they are not preferentially selected; when they do disclose, human selectors are initially biased against them but progressively learn to prefer them over human partners. We find little evidence that human candidates adapt their behaviour in response to AI competition — illustrating how repeated interaction with AI can reshape social beliefs, decisions, and interaction structures in hybrid societies.

Talk 2 — How knowledge of social network structure shapes strategic gossip (Alice Xia)

Gossip is one of the most common ways information travels through social groups, yet sharing it often requires strategic behaviour: the same web of connections that lets a message spread also creates the possibility it reaches the wrong person. Across controlled experiments and a large real-world network, we find that people rely on distance and popularity to share gossip strategically. Computational modelling shows behaviour is best explained by mental representations of the network that leverage multistep connectivity.