Picture-Evoked Affect and Review Helpfulness: An Affect-as-Information Perspective

Abstract

Viewing pictures evokes affective feelings and influences perceptions. Although pictures in online reviews shape readers’ affect, including valence and arousal, how picture-evoked affect impacts reader perceptions of review helpfulness remains understudied. Based on affect-as-information theory, we propose that picture-evoked valence exhibits a positive effect on perceived review helpfulness, and picture-evoked arousal exhibits a U-shaped effect. Perceived processing fluency mediates the effect of picture-evoked valence, and perceived review credibility mediates the effect of picture-evoked arousal. Text-evoked valence positively moderates the effect of picture-evoked valence, whereas text-evoked arousal negatively moderates the effect of picture-evoked arousal. A laboratory experiment and a large-scale field test support our hypotheses. We further show that picture-evoked affect is associated with various interpretable picture features. Our empirical strategy involves techniques of computer vision, deep learning, and econometrics. Our work deepens the understanding of visual and affective ingredients of helpful reviews, derives theoretical insight into underlying mechanisms and moderating effects, and offers practical implications for online review platform design and online reputation management.

Keywords: Affect-as-information, helpfulness, pictures, valence, arousal, perceived processing fluency, perceived review credibility

JEL Classification: M15, M21, M31