Learning to Trust: Cognitive Mechanisms in Human鈥揂I Cooperation
Recent studies suggest that humans are often distrusting and less willing to cooperate with artificial agents, even when such cooperation would yield better outcomes. At the same time, other research presents contradictory evidence, indicating that people may sometimes overtrust algorithms. However, the underlying reasons for this cooperative trust or distrust remain unclear.
This PhD research aims to investigate trust in human-AI cooperation by examining the cognitive processes that shape cooperative behaviour between humans and artificial agents. Using social dilemma games, we explore how trust is formed and dynamically learned through repeated interactions with AI systems, particularly in response to feedback. Specifically, we examine what initial expectations people hold about algorithmic partners and how they update these expectations when they are either confirmed or contradicted by the AI鈥檚 actual behaviour over time. In addition, we employ computational modeling approaches (e.g., reinforcement learning drift diffusion models) to identify the cognitive processes that underpin this learning.
Involved researchers
- I am a PhD candidate in Psychology. My doctoral research explores how trust is formed and dynamically adjusted through repeated interactions with AI systems, particularly in response to feedback
- I am an associate professor with a special interest in social and cognitive aspects of Human-AI interactions. I am co-directing the GoalLab with Ruud Custers
- I am a professor in Psychology, interested in human habits, goals, and autonomy. I am involved in several projects studying basic and applied questions, such as intentionality, Human-AI interactions and health
Funding
- Werkdrukimpuls fonds