Usability Assessment of Conversational Agents in Healthcare: A Literature Review

Conversational agents (CA) are chatbot-based systems supporting the interaction with users through text, speech, or other modalities. They are used in an increasing number of medical use cases. Even though usability is considered a prerequisite for the success of mHealth apps using CA, there is still no standard procedure to study usability of health CA. In this paper, we report the results from a systematic literature review aiming at identifying study designs, tools, and metrics used to assess usability in health CA. We searched three bibliographic databases (PubMed, Scopus, IEEE Xplore) for papers reporting on CA in healthcare to extract information on the usability assessment of those CA. From 273 retrieved results, we included 66 papers for full text review. 34 of them reported on usability assessments. A broad range of tools is used (e.g. SUS, UEQ), but also individual questionnaires are exploited. The examined studies use scenario-based setups but assess also real-world usage. Exploratory setups are rarely reported. Due to the differences in the study designs and assessment tools, it is impossible to compare usability among CA. Thus, we recommend to develop a standardised procedure that can be always applied and which can be enriched by assessments needed for evaluating usability of CA-specific features.