Decoding Trust in Conversational Agent Interactions

Conversational agents are used as mediators between public organisations and citizens, necessitating a better understanding of how trust is affected when such technology is implemented. These agents have characteristics related to trust, called trust markers, which can increase or decrease trust. To investigate these trust markers, we analysed conversational data between citizens and a public organisation. From this data, we identified a set of trust markers. Preliminary findings show that user feedback and switching to a human agent are important trust markers, alongside the duration and length of the conversation. This work-in-progress examines the relationships between these trust markers and other potential variables. By uncovering these relationships, we can pinpoint where trust is compromised, allowing a human agent to intervene and prevent the erosion of trust between citizens and organisations.