How Children Learn from People: Child Development and Model Rearing

​How do children learn when to explore, whom to trust, and how to respond to the world around them? Early development is deeply social: infants differ in temperament, learn through repeated interactions with caregivers, and gradually develop ways of regulating emotions and navigating their environments.

​In this talk, Xi Jia Zhou will introduce key ideas from developmental psychology, focusing on early social and emotional development, temperament, attachment, and caregiver–child interactions. Drawing on observational research and computational models of infant learning, she will explore how a learner's characteristics and social experiences jointly shape development and behavior.

​We'll then turn to an emerging question: What, if anything, can child development teach us about how AI models learn from human feedback and social environments?

​Xi will briefly introduce her ongoing thinking on model rearing and developmental AI safety, opening the floor for discussion: Where might these parallels be useful? Where do they break down? And what new, testable research questions could they inspire?

​This will be an accessible, discussion-oriented session. No prior background in developmental psychology is needed. We welcome perspectives from AI safety, machine learning, cognitive science, and beyond!

​About the Speaker

​Xi Jia Zhou is a PhD candidate at Stanford University, specializing in developmental psychology and learning sciences, and an AI safety researcher. Her work spans naturalistic infant–caregiver interactions, computational models of social learning, and how human feedback shapes AI behavior.