Abstract
Eye gaze is a powerful directional cue, and previous research has found that when a centrally-presented face looks away from the imminent location of a target, judgements about that target are slower and more error prone, and the face is less likely to be selected as trustworthy than another face that cues location validly. Previous research into this area (Bayliss & Tipper, 2006) has used a 2-alternative forced choice (2afc) selection when gauging trustworthiness, which cannot tease apart exactly how these helpful or misleading gaze behaviours change apparent trustworthiness. In three of four experiments, we use two scalar ratings of trustworthiness – one before and one after trust induction – to measure how apparent trustworthiness changes based on faces’ gaze behaviour. In Experiment 1, we show that the trust effect for neutral faces is driven by a unidirectional decrease for misleading faces, while trustworthiness ratings for valid-cueing faces remain constant across the experiment. In Experiment 2, we show that this effect is specific to trustworthiness, and the pattern does not generalise to ratings of likeability, meaning that the social inferences being made likely do not reflect a simple valence judgement. In Experiment 3 we replaced the rating scales with two morphed face images (more/less trustworthy) and asked participants to select which they had seen in the experiment. Cueing validity did not affect which face was chosen, indicating that the effect does not interfere with face memory. Finally, in Experiment 4, we show that changing the expression of the faces in the original gaze-cueing paradigm changes the nature of the trustworthiness ratings: trust effects for smiling faces show the decrease of trust for misleading faces that neutral faces show, but also an increase for valid-cueing faces. Taken together these findings hint at a complex online system for updating social impressions during interactions.
Meeting abstract presented at VSS 2015