Metacognitive challenges in Human-AI interaction

Human–AI interaction is as much about how people think as about what AI does. Expectations and self-monitoring shape how users engage with intelligent systems and interpret their own performance. Our research shows that merely believing in AI support can enhance both confidence and task performance, even when no real assistance is provided— a placebo effect of AI. This optimism proves remarkably resilient: even when AI is described as unreliable or stress-inducing, people continue to expect improvement and improve their task performance, demonstrating the robustness of the effect. More recent work reveals a new paradox. While AI reliably enhances cognitive performance, it simultaneously impairs users' insight into their own task accuracy. People tend to overestimate their competence, and those with greater AI literacy are often the most overconfident. Taken together, these findings suggest that the next challenge for human-centred Human–AI interaction is to design for human metacognition, to help users remain aware of their interaction with AI.

About the Speaker

Dr Robin Welsch

Assistant Professor of Engineering Psychology at Aalto University

Dr. Robin Welsch is an Assistant Professor of Engineering Psychology at Aalto University, Finland, where he investigates human-computer interaction, virtual reality, and artificial intelligence. His research explores how people interact with emerging technologies, including notable work on the "AI ghostwriter effect"—examining user perceptions of AI-generated content—and societal attitudes toward human augmentation technologies. He has published in leading venues such as ACM Transactions on Computer-Human Interaction and the CHI Conference on Human Factors in Computing Systems, with research spanning topics from the placebo effect of AI to proxemics in augmented reality and social interactions in virtual environments.