The Impact of User Momentary Emotional State on Trust in a Faulty Chatbot

Abstract

Chatbots have become increasingly prevalent in our daily lives. Research shows that, while some users trust such technology even with sensitive information, others refuse to rely on this technology for support. When explaining trust in such technology, existing human-technology trust approaches are overlooking factors like users’ emotional states. This three-part experimental study in a real-world setting examines the influence emotional states have on trusting a faulty chatbot. Participants interacted with a chatbot in part one of the study to schedule an appointment some days later for part two. Upon arrival for part two, participants were informed of a mistake in scheduling their appointment. The experimental manipulation consisted of fault attribution to the chatbot in one group, whereas no attribution to fault in the other group. In part three, participants then chose to either schedule a new appointment using the same chatbot or email.A total of N = 58 participants participated in the first part of the study, while a total of N = 30 participants completed the study. The main finding indicates that a more positive momentary emotional state towards the chatbot was related to higher self-reported trust, even after the chatbot made a mistake. However, trust did not affect trusting behavior afterwards. Considering these findings, we contribute theoretical advances to the existing trust research landscape in a setting relevant for everyday life. We also discuss potential explanations for the resulting pattern of effects and implications on chatbot design for the role of emotional states when trusting chatbots.

Keywords: chatbot, Trust, emotional state, affinity for technology, fault attribution, trusting behavior