Recommendation agents (RAs) are regularly used in online environments to give consumers advice on products. Since social components of human-like RAs (humanoid avatars) are important components in their adoption and use, this study focuses on how the design of the anthropomorphic interface of RAs in terms of social demographics, namely ethnicity and gender, can enhance the RA’s social presence to facilitate their adoption. Since social presence has been shown in the literature to predict the adoption and use of RAs, we examine whether match or mismatch in terms of the anthropomorphic RA’s ethnicity and gender can enhance the user’s social interaction with an RA.

To overcome concerns of social desirability bias and political correctness when users assess the social presence of RAs that vary in their ethnicity and gender, we conducted a functional Magnetic Resonance Imaging (fMRI) study to complement a traditional behavioral experiment. Our goal was to explain prior behavioral findings that showed that ethnicity (as opposed to gender) match is associated with higher social presence, particularly among women. Specifically, brain activity was captured in an fMRI scanner while users who varied on their ethnicity and gender to either match or mismatch the ethnicity and gender of four RAs evaluated each of the RAs on their social presence.

Besides contributing to the neuroscience literature by identifying the brain activations that relate to social presence, the fMRI results shed light on the nature of social presence and explain earlier behavioral findings by showing gender differences in the neural correlates of social presence in terms of ethnicity and gender match and mismatch. Implications on designing anthropomorphic interfaces to embody social demographics to enhance social presence are discussed.