Conversational Agents, Natural Language Processing, and Machine Learning for Psychotherapy

Mental illness is one of the most pressing public health issues of our time. Economic constraints, social stigma, and scarce availability of professionals require, on one hand, to augment clinical support and quality, and on the other hand, to create instruments able to augment the treatments and to enrich training and supervision methods. The use of Conversational Agents (CAs) in the field of mental health and psychotherapy is in the early stages of development. This chapter helps to understand the reasons of this delay and suggests some strategies that could be adopted to compensate for it. The chapter analyzes and discusses the most recent literature dealing with CAs, Natural Language Processing (NLP), and Machine Learning (ML) supporting the mental health domain, with a special focus on psychotherapy and psychological support, evaluation of strategies and systems, analysis of clinical interactions, and training and supervision of professionals. Moreover, the chapter discusses several models available as research frameworks, prototypical solutions, or commercial systems. The aim of the proposed reasoned survey is twofold: to allow the enrichment of clinical interventions, and to invite NLP and ML experts to consider this applicative domain as particularly important for the development of CAs. Human-like interaction – imitated or complemented by CAs – is a central aspect of this chapter. In particular, the relevance is clear of empathic-oriented behaviors, sentiments (and emotion analysis), prosodic and “mirroring” competencies, in the context of evidence-based protocols that try to ensure the efficacy of augmented psychotherapeutic strategies and the subjects' and therapists' adherence to them.