Abstract
Conversational facial expressions are the most pervasive forms of facial expressions in real social contexts (e.g., Rozin & Cohen, 2003), and used to manipulate the flow of conversation – for example, showing encouragement can extend interactions, whereas showing doubtful can re-route or terminate it. Although conversational facial expressions play a central role in human-human interaction (e.g., Bavelas & Chovil, 2000; Chovil, 1991) and human-robot interaction (e.g., Cassell, 2000), comparatively little is known about their face movement patterns, and whether these patterns are similar across cultures (but see also Ekman, 1979; Nusseck, Cunningham, Wallraven, & Bülthoff, 2008). Here, we address this knowledge gap by modelling 50+ dynamic conversational facial expressions in two cultures (54 Western, 58 East Asian observers) using a facial expression generator (Yu, Garrod, & Schyns, 2012), reverse correlation (Ahumada & Lovell, 1971) and subjective perception (see also Gill, Garrod, Jack, & Schyns, 2014; R. E. Jack, Garrod, & Schyns, 2014; R. E. Jack, Garrod, Yu, Caldara, & Schyns, 2012). Cross-cultural comparison of the resulting dynamic facial expression models showed clear cultural similarities in facial expressions such as contented, offended, and sorry that correspond to culturally common facial expressions of emotion (see R. Jack, Sun, Delis, Garrod, & Schyns, 2016). In contrast, facial expressions such as doubtful, sympathetic, and indecisive showed culture-specific accents. Together, our results enhance knowledge of conversational facial expressions, and anticipate their application in informing the design of culturally aware digital economy technologies, such as social robots (e.g., Foster et al., 2012) and virtual humans (e.g., Swartout et al., 2006) to support the evolving communication needs of modern society.
Meeting abstract presented at VSS 2017