Time: 12:30 PM, September 17, 2026 (Thursday)
Location: Room 1100, 11th Floor, Lyu Dalong Building, Tsinghua University
Presentation Topic:
Influence of Human-AI Agent Social Interaction on Food Choices and Psychological Mechanisms
Presenter: Xiyu Guo
Abstract：
Previous studies have confirmed that interpersonal interaction shows a positive impact on improving sustainable food choices. With the development of artificial intelligence technology, intelligent agents are increasingly intervening in human decision-making processes as “social actors.” However, it remains unclear whether social interaction between humans and intelligent agents can similarly promote Chinese consumers’ sustainable food choices. We conducted three studies to investigate the impact of human-AI agent social interaction on Chinese consumers’ food choices and its psychological mechanisms, while also examining the boundary conditions and ecological validity of this effect.
Study 1 showed that health-related rather than hedonic-related social interaction with an intelligent agent effectively prompted participants to make more sustainable food choices, which manifested as a significant reduction in the choice of meat-heavy meals after the interaction compared to before. This effect remained robust after controlling for variables such as genders, levels of autonomy, health or hedonic motivations, and the duration or number of interaction rounds.
Study 2 found that the interaction content itself, rather than the persona label of the intelligent agent, drove this effect. ERP results further revealed that this effect operated through an informational social influence pathway, with individuals’ food choices exhibiting a dynamic processing pattern from early-stage cognitive conflict monitoring (N2) to mid-to-late-stage meaning integration (N400).
Study 3 indicated that the effect was not significantly moderated by participants’ age or the agent’s degree of humanization. Furthermore, the positive effect of human-AI agent social interaction remained robust in both complex decision-making scenarios involving numerous food options and real-world eating contexts, but its duration may be relatively short-lived.
In summary, this research demonstrates that individuals engage in deep cognitive processing of the interaction content during human-AI agent social interactions and alter their food choices under the influence of informational social influence. This effect is substantiated by neural evidence and characterized by robustness, universality, real-world generalizability, and time-sensitivity. This research not only illuminates the pivotal driving role of informational content in human-AI agent social interactions and advances the understanding of how non-interpersonal social influence operates at the cognitive neural level, but also provides scientific foundations and design guidelines for developing low-cost, highly universal digital dietary intervention systems.
