Testing Conversational Agents as a Digital Companion

Overview

There is a need and opportunity to improve the supports, transitions, and life outcomes of people with autism spectrum disorder. Compared to their neurotypical peers, autistic teenagers and adults report poorer mental health and quality of life and have higher rates of unemployment or underemployment and low participation in post-secondary education. Nearly 40% spend little or no time with friends. Although autism awareness has grown considerably in recent decades, much more can be done to improve the life outcomes for people with autism. Cost-effective, affordable and scalable support systems are needed as well as ongoing assessments and personalized support plans that focus on individual strengths and challenges in different contexts (college, work, community life) across the life span. This requires adaptive interventions and regular consultation with and between stakeholders. It also requires a rigorous approach to measuring outcomes that are not one-size-fits-all and do not expect everyone to reach, or have, the same goals. To meet these needs, the investigators leverage an already successful technology platform with two conversational-relational agents to be a digital companion and coach to autistic young adults (AYA, ages 18 to 35 years). The technology will be used to scale a strong theoretical and conceptual approach that has proven successful in meeting the individual needs and personalized outcomes of autistic students through a collaborative consultation model for promoting competency and success (COMPASS) combined with Goal Attainment Scaling (GAS). To deliver personalized coaching, empathy, and outcomes at scale, GAS/COMPASS was translated into software-driven evidence-based coaching protocols in collaboration with clinical, academic, and community partners. In this study, the digital coaching program and all research protocols are pilot tested in a 10-week experiential trial with AYA.

There is a need and opportunity to improve the supports, transitions, and life outcomes of people with autism spectrum disorder (ASD). Autism affects about 5.4 million adults in America (2.2% or 1 in 45 adults) and at least 78 million people and their families worldwide.1,2 Estimates for children are similar (1 in 44). Classified as a neurodevelopmental disorder, autism typically manifests in early childhood before the age of 3 years and affects social and communication skills, behavior, and outcomes throughout the life span. Teenagers and adults with autism have higher rates of unemployment or underemployment, low participation in post-secondary education, and nearly 40% spend little or no time with friends. Although awareness and scientific interest has grown significantly over the past 20 years, many children and adults are not diagnosed until later in life and much more can be done to improve the life outcomes for people with autism. Large knowledge gaps exist, such as what support strategies are effective for whom and when, and which ones have long-term and meaningful impact. Essentially, the mechanisms of positive change in autism, both behavioral and neurobiological, are largely unknown. Because autism has major implications for lifelong outcomes, cost-effective, affordable and scalable support systems are key, yet missing for many individuals and families. Given that autism is a complex, heterogenous developmental condition that changes with and affects development, The Lancet Commission on the future of care and clinical research in autism recently stated that single assessments and treatments are never sufficient in ASD. Ongoing assessments and personalized supports that focus on individual strengths and challenges in different contexts across the life span are needed. Meanwhile, autistic people, their family members and community organizations have called for greater emphasis on enhancing quality of life (QOL). The World Health Organization (WHO) defines QOL as "an individual's perception of their position in life in the context of the culture and value systems in which they live and in relation to their goals, expectations, standards, and concerns". In response, investigators at the Center for Autism Research Center at Children's Hospital of Philadelphia leveraged the patient reported outcome measurement information system (PROMIS) developed by the NIH to comprehensively assess QOL in autism, across age groups and genders. The PROMIS Autism Battery (PAB) includes 18 self-report scales for adults that show profound differences between autistic and neurotypical adult samples across virtually all PAB domains, including subjective wellbeing (SWB), relationships (including supports, companionship and social isolation), emotional distress (including anxiety and depression), health (including sleep and mental health), and adulthood (including self-efficacy). In response to these needs and to NIMH's special topic of interest to "Develop and test new and augment existing digital health interventions that are personalized, engaging, adaptive, sufficiently challenging, and optimal for maximizing real world functional improvements," Friendi.fi Corporation (Friendi.fi) is leveraging its AI-driven technology platform and conversational agent that works on mobile phones through standard SMS (Short Message Service) texting, to be a mobile digital companion to autistic people in early adulthood. The platform and agent is used to scale a strong theoretical and conceptual approach that has proven successful in meeting the individual needs and targeted outcomes of autistic students through a collaborative model for promoting competence and success (COMPASS) combined with Goal Attainment Scaling (GAS). COMPASS considers the challenges and strengths of the individual and environment, as well as the balance between them. Critically, the role of the environment as an enabler (support) or barrier (challenge) to success lowers the burden on autistic individuals as a source of dysfunction. In this model, technology is viewed as an environmental support that builds on the strengths, needs, and goals of the individual. AYA will be enrolled in a 10-week experiential trial using the novel digital coaching program and agent. The trial serves several purposes: (1) It will test the preliminary efficacy of the digital intervention by assessing within-person change over time. The trial is powered to detect a minimal important change (MIC) in the proposed standardized and normalized PAB measures; (2) The observed effect size will be used to finalize the sample size calculation for a properly-powered future randomized control trial (RCT); (3) By including a control group, the feasibility of all clinical and study protocols is tested prior to implementing the RCT. Twenty-six AYA will receive daily coaching via the digital agents (AGENT group) and twelve will receive traditional coaching using face-to-face meetings with human counselors (CONTROL group). In both groups, intervention delivery and patient-reported outcomes data are collected.