As artificial intelligence (AI) becomes increasingly capable of supporting complex decisions, organizations face a fundamental question: When should humans move beyond collaborating with AI to delegating decision authority to it? Although prior research has extensively examined whether individuals follow AI recommendations, we know considerably less about when they are willing to relinquish control altogether and whether they appropriately adjust such decisions as AI capabilities change. We investigate this question in a controlled experiment in which participants perform a repeated strategic search task, initially collaborating with an AI advisor before deciding whether to delegate subsequent decisions entirely to the AI. We experimentally vary AI performance, including consistently high and low quality, sudden performance deterioration, and gradual improvements or declines. We further distinguish between delegation decisions that are binding for the remainder of the task and those that allow participants to periodically reconsider and reclaim control. This design enables us to examine how individuals learn about AI capabilities, translate their experience into automation decisions, and respond to changes in AI reliability. By distinguishing the decision to follow AI advice from the decision to delegate decision authority, our study advances understanding of the behavioral foundations of AI automation and the conditions under which human oversight adapts to the evolving capabilities of AI.

Cognitive neuroscientists have a societal obligation not only to generate knowledge, but to disseminate it and steward its responsible application. This seminar brings together two researchers whose work spans the arc from fundamental neuroscience to real-world implementation, addressing the successes and challenges of translating findings into practice. Katia Steinfeld (UniL) connects her research on how myopia shapes visual and multisensory development to Escolhares, the non-profit she founded that delivers eye care to over 4,000 children annually in underserved Brazilian communities. Meike Ramon (BFH) draws on two decades of research into the neural and cognitive mechanisms of face identity processing to trace her path from basic questions about how the brain recognizes faces to applied collaborations with the public sector—spanning clinical assessment, security and identity verification, and everyday trust and social interaction. Together, these talks offer complementary perspectives—one from an established research trajectory, one from an early-career scientist bridging lab and field—on what it takes to move cognitive neuroscience out of the laboratory and into domains such as education, public policy, and security. Both speakers reflect on balancing scientific rigor with real-world relevance, and on building the dialogue with policymakers and practitioners needed to responsibly translate research into societal impact.