The Social Transfer Function: How Dynamic Predictions of Facial Consequences Drive Judgements of Social Contingency

Abstract

Human social interactions display a complex choreography of dynamic, non-verbal signals such as nods and blinks. Yet, little is known about how these cues contribute to the detection of social contingency, i.e., how exactly does one know that two people are interacting with one another? To answer this question, we developed an experimental paradigm where observers had to recognize genuine and fake contingency in naturalistic, dyadic interactions. With two independent experiments (N = 206), we show that perception of contingency is robust even with severely degraded interactional stimuli and that observers causally rely on the link between a speaker’s speech and the listener’s motion. Crucially, inference about appropriate contingency is driven by temporally aligned, dynamic predictions rather than average quantity of movement or any lagged correlation between speech and motion. To model this perceptual process, we introduce the concept of a ‘social transfer function’: a pre-learned schema that generates predictions of a listener’s speech-contingent response and, based on how well-matched the predictions are to observed behaviour, is able to predict observer judgments of genuine contingency. Taken together, these results provide mechanistic insights into the audiovisual features that contribute to perception of social contingency, and open new ways to identify markers of contingency in people with disorders of consciousness, autism and social anxiety.