Abstract
The use of chatbots in healthcare is an area of study receiving increased academic interest. As the knowledge base grows, the granularity in the level of research is being refined. There is now more targeted work in specific areas of healthcare, for example, chatbots for anxiety and depression, cancer care, and pregnancy support. The aim of this paper is to systematically review and summarize the research conducted on the use of chatbots in the field of addiction, specifically the use of chatbots as supportive agents for those who suffer from a substance use disorder (SUD). A systematic search of scholarly databases using the broad search criteria of ("drug" OR "alcohol" OR "substance") AND ("addiction" OR "dependence" OR "misuse" OR "disorder" OR "abuse" OR harm*) AND ("chatbot" OR "bot" OR "conversational agent") with an additional clause applied of "publication date" ≥ January 01, 2016 AND "publication date" ≤ March 27, 2022, identified papers for screening. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines were used to evaluate eligibility for inclusion in the study, and the Mixed Methods Appraisal Tool was employed to assess the quality of the papers. The search and screening process identified six papers for full review, two quantitative studies, three qualitative, and one mixed methods. The two quantitative papers considered an adaptation to an existing mental health chatbot to increase its scope to provide support for SUD. The mixed methods study looked at the efficacy of employing a bespoke chatbot as an intervention for harmful alcohol use. Of the qualitative studies, one used thematic analysis to gauge inputs from potential users, and service professionals, on the use of chatbots in the field of addiction, based on existing knowledge, and envisaged solutions. The remaining two were useability studies, one of which focussed on how prominent chatbots, such as Amazon Alexa, Apple Siri, and Google Assistant can support people with an SUD and the other on the possibility of delivering a chatbot for opioid-addicted patients that is driven by existing big data. The corpus of research in this field is limited, and given the quality of the papers reviewed, it is suggested more research is needed to report on the usefulness of chatbots in this area with greater confidence. Two of the papers reported a reduction in substance use in those who participated in the study. While this is a favourable finding in support of using chatbots in this field, a strong message of caution must be conveyed insofar as expert input is needed to safely leverage existing data, such as big data from social media, or that which is accessed by prevalent market leading chatbots. Without this, serious failings like those highlighted within this review mean chatbots can do more harm than good to their intended audience. [Abstract copyright: © 2022 S. Karger AG, Basel.]