Enhancing Empathy in Text-Based Teletherapy with Emotional State Inference

Abstract

Over half of the U.S. population lives in an area without adequate access to mental health care and the unmet demand for mental health services has shifted to care providers who have not been trained to provide mental health support. This work represents a step toward addressing this supply-demand imbalance by applying recent advances in conversational AI. The central hypothesis of this work is that both the quality and efficiency of text-based telehealth can be improved through recent advances in conversational AI. This hypothesis was evaluated with three aims: (Aim 1) explored the ability of computational methods to infer high-fidelity representations of emotional states as a precursor to empathy, (Aim 2) evaluated these representations as features for a transformer-based empathic response predictor, (Aim 3) piloted this system as a component of a teletherapy platform for the delivery of problem-solving therapy by nurses and psychologists. The results of these aims validate this core hypothesis by successfully collecting emotional health information through an automated SMS-based intervention and by significantly improving empathic accuracy and reducing response times of human care providers using an AI-augmented chat interface. Together the components of this dissertation provide a unified solution that can help to increase access to mental health care by automating the remote monitoring of emotional health, expanding the number of individuals who can provide protocolized care, and enhancing the efficiency and empathy of the care provided. During the course of this work, I developed a novel evaluation paradigm to better measure how emotion recognition systems can help to track emotional health through automated journaling exercises, applied these measures to predict empathic responses, and evaluated a support tool to assist care providers in delivering problem-solving therapy.