Users Perceptions and Acceptances of Text-based Therapist Conversational Agents

Abstract

As there has been a surge of interest and the development of services to supplement existing mental health interventions via artificially intelligent assistants, therapist conversational agents have received attention as a promising area. Recently, commercially available therapist conversational agents on mobile messaging apps were already being widely used to help users keep track of their feelings, thoughts, emotions, and problems in monitoring their mental health.

According to empirical research in the human-computer interaction area, when users expectations are not satisfied with, users level of trust in technology decreases, which also drops users level of acceptances. Since conversation is related to a variety of expectations and social cues especially in counselling settings, designing a conversational user interface that meets users expectations and their desired perceptions is complicated. Since these therapist conversational agents have recently become commercially available on mobile messaging apps, very little research has been done on users expectations or acceptances of text-based therapist conversational agents on mobile messaging apps.

With this in mind, we will explore users expectations, perceptions, and acceptance of therapist conversational agents. To tailor users expectations and preconceptions toward therapist conversational agents on mobile settings, it is crucial to examine what are the important elements of a graphical and conversational user interface in advance. Further, we examine the perceptions that users have after interacting with a therapist conversational agent. We finally explore how users perceptions influence user acceptance and adoption of therapist conversational agents.

On the interface level, agents anthropomorphic features critically affect users perceptions and impressions on them. Often, indiscriminate anthropomorphic features caused negative impressions on users perceptions toward agents. On the perception level, the most significant users perceptions of acceptance turned out to be perceived intelligence. This result implies that users do not expect the technology to only work passively, according to their command

users expect conversational agents to be more natural, conscious, and competent, as if they are human assistants. Agents level of social support and self-disclosure, which came out as determining factors on users acceptances, can be the catalyst that contributes to building a relationship between artificial intelligence based conversational agents and human users.

By adapting users anthropomorphic preferences to the conversational and graphical interfaces, we expect to improve the users impressions and acceptances about conversational therapist agents. The findings resulting from this study will help designers and researchers achieve a better understanding of users experiences, perceptions, and acceptances on therapist conversational agents.