Do I Always Want You to Smile? How the Smile of Computer Agents Is Not Always Perceived Positively

Abstract

Although smiles are typically interpreted as positive social cues, their meaning can shift depending on situational and contextual factors. In particular, in environments characterized by high social risk or perceived corruption, a smile may be interpreted as manipulative or deceptive rather than trustworthy. Such context-sensitive interpretations are well-documented in human-human interaction (HHI) and are increasingly relevant in human-agent interaction (HAI), where social norms are often transferred to artificial agents. This study investigates whether these nuanced interpretations of smiles also apply to interactions with virtual agents. But does that mean it is better not to smile in the corrupted community? For this, we conducted a study with two conditions: Low-Risk and High-Risk. Our findings suggest that regardless of the expression, the agents are evaluated more positively in low-risk situations than in high-risk conditions. This study contributed to understanding the importance of the functionality of context in HCI, i.e., where the risk is accompanied.