Age-related patterns in the creation of social prediction models: a study using immersive virtual reality.

Exposure to environmental regularities allows individuals to build internal models, supporting the integration of prior expectations with sensory input in social contexts. However, how social prediction models develop across age and experience remains unclear. To address this, we investigated whether the formation of social predictive models varies across age and experiences in neurotypical individuals (early adolescents, young adults, and adults). Participants completed, in counterbalanced order, two scenarios of an immersive virtual reality (IVR) task that facilitates the creation of top-down predictions during action perception, requiring the participants to implicitly associate the hand movement of four avatars toward one preferred object among three. Self-report questionnaires and standard social cognition tests were used to assess embodiment, presence, usability in the IVR setting, and the direct association between the novel IVR task and social-cognitive tests. Results showed significant performance improvement from the first to the second scenario in all groups, indicating learning and refinement of predictive models. Early adolescents performed significantly worse and showed a lower learning rate than adult groups, who did not differ from each other, suggesting slower formation of social prior expectations in early adolescents. The IVR paradigm was feasible across all age groups and proved valid for assessing social processing. These findings support a developmental trajectory in which the creation of predictive internal models improves with age and experience.