Do young children learn better from an informative or affiliative virtual agent?

Young children are active communicators well before they become proficient speakers. Even in preverbal stages, they engage in complex turn-taking behaviors, and their development is shaped by the responsiveness of their social environment. Yet not all responses are equal: caregivers often switch between affiliative responses (e.g., emotional mirroring, encouragement) and informative responses (e.g., labeling, correction). The relative impact of these communicative styles on early learning and engagement remains surprisingly underexplored. This internship will investigate how different styles of social feedback from a virtual agent influence young children's communicative behaviors. Students will contribute to the development of an interactive agent capable of producing either affiliative or informative responses, potentially using generative AI or scripted interactions. We will then experimentally evaluate how these differing interaction styles shape children’s vocalizations, gaze behavior, and/or learning outcomes. The project offers hands-on experience at the intersection of developmental psychology, human-computer interaction, and AI. Depending on interest and background, students may contribute to: Designing and prototyping the virtual agent’s behavior Experiment design and study with child participants Annotating and analyzing video/audio recordings of child-agent interaction Theoretical integration with frameworks of language acquisition, social learning, and joint attention This work has implications for both cognitive science theory and the design of socially intelligent agents in educational settings.