Abstract

Delivering bad news, such as a rejection, can be very difficult for both the deliverer and the recipient. In particular, the bearer is at risk of being blamed by the recipient. However, what happens when the deliverer of bad news is a conversational agent (CA)? Moreover, how does a humanlike CA design influence influence the user's attribution of blame? We conducted an online experiment with 101 participants to investigate how perceived humanness in CAs that deliver bad news influences the user's attribution of blame. Our results indicate that perceived humanness has contradictory effects on the attribution of blame, increasing and decreasing it simultaneously. Based on our findings, we suggest that designers of CAs that deliver bad news should be very concerned with humanlike design, as their well-intentioned design choices can have unintended consequences.

Share

COinS