Human adults can figure out what happened by combining evidence from different sensory modalities, such as vision and sound. How does the ability to integrate multi-modal information develop in early childhood? Inspired by prior computational work and behavioral studies with adults, we examined 3- to 8-year-old children's ability to reason about the physical trajectory of a ball that was dropped into an occluded Plinko box. Children had to infer in which one of three holes the ball was dropped based on visual information (i.e., where the ball landed) and auditory information (i.e., the sounds of the ball colliding with parts of the box). We compare children's responses to the predictions of four computational models. The results suggest that although even the youngest children make systematic judgments rather than randomly guessing, children's ability to integrate visual and auditory evidence continues to develop into late childhood.

Conversational AI devices are increasingly present in our lives and even used by children to ask questions, play, and learn. These entities not only blur the line between objects and agents—they are speakers (objects) that respond to speech and engage in conversations (agents)—but also operate differently from humans. Here we use a variant of a classic false-belief task to explore adults' and children's attributions of mental states to conversational AI versus human agents. While adults understood that two conversational AI devices, unlike two human agents, may share the same "beliefs" (Exp.1), 3- to 8-year-old children treated two conversational AI devices just like human agents (Exp.2); by 5 years of age, they expected the two devices to maintain separate beliefs rather than share the same belief, with hints of developmental change. Our results suggest that children initially rely on their understanding of agents to make sense of conversational AI.