Pai.ACT - An Artifical Intellegience Driven Chatbot Assisted ACT

Pai.ACT: A Deep-learning-powered, Smartphone-delivered, Acceptance and Commitment Therapy for Parents of Children With Special Needs: A Feasibility Study

Hong Kong Christian Service, Hong Kong Young Women's Christian Association, Yang Memorial Methodist Social Service, Hong Kong Federation of Youth Groups, The Pamela Youde Nethersole Eastern Hospital

Limited psychological support for parents of children with special needs in Hong Kong can profoundly impact the child rehabilitation process and the well-being of parent-child dyads. Leveraging previous evidence from our team's research, we have developed Pai.ACT, the first deep learning-based mental health advisory system for parents. Pai.ACT incorporates the counselling logic of Acceptance and Commitment Therapy (ACT) through natural language processing, enabling parents to engage in human-like voice-to-text conversations and receive assessments and stepped-care mental health interventions, including guided self-help materials and real-time, individual-based counselling based on ACT. Following the research and development phases, we aim to kick off the utilisation of Pai.ACT by (1) pilot-testing its feasibility, acceptability, and potential efficacy in improving mental health outcomes for parents of children with special needs and (2) researching to determine the most optimal service model for parents by exploring their perceptions through focus group interviews. Pai.ACT offers accessible and comprehensive mental health services to all Chinese-speaking parents, addressing their psychological burden in caring for children with special needs. Pai.ACT could bring substantial and enduring societal benefits to Chinese-speaking families by integrating mental health support services for family caregivers with current child rehabilitation services and non-governmental organisations. Furthermore, this could contribute to reducing the public stigma attached to special needs children while increasing mental health awareness.

INTRODUCTION Children classified as having special needs encompass those with chronic developmental, behavioural, and emotional conditions requiring specialised and extended health services beyond those required by children in general. Common types of special needs in Hong Kong include autism spectrum disorders (ASD), attention-deficit hyperactivity disorders (ADHD), and developmental disabilities or delays. Notably, there has been a significant increase in the number of children with special needs in the region. In 2019, the Child Assessment Centre under the Department of Health reported a rise in newly diagnosed cases and referrals for specialty follow-up services for preschool-aged children, with annual growth ranging from 6.2% to 10.7%. On a broader scale, Hong Kong offers a range of health and social services for families of children with special needs, but they are marked by prolonged waiting times for multidisciplinary assessments, diagnoses, and subsequent intensive rehabilitation services. This extended waiting period not only delays crucial support but also intensifies the psychological burden experienced by parents. Without a certified diagnosis from mental health professionals, parents face significant challenges in accessing public health services for their children's rehabilitation. The local mental health landscape predominantly focuses on severe mental illness, prioritising the children's needs over the psychological well-being of parents. Accessing mental health services becomes even more challenging due to caregiving demands, time constraints, geographic limitations, and a scarcity of providers with expertise in caring for special needs children. Self-stigma further impedes parents from seeking psychotherapeutic support, despite recognising its impact on their well-being and parenting behaviours. Notably, parents of special needs children exhibit higher levels of parenting stress and depressive symptoms compared to parents without special needs. Acceptance and Commitment Therapy Acceptance and Commitment Therapy (ACT) has recently emerged as a promising, low-intense psychotherapy for addressing parenting challenges. By practising core ACT skills like cognitive defusion, acceptance, and value clarification, parents can better navigate difficult parenting experiences and reconnect with their caregiving qualities. Extensive research supports the effectiveness of ACT in improving psychological well-being and reducing somatic complaints across diverse populations. Recent clinical trials targeting paediatric conditions such as asthma, ASD cerebral palsy, and acquired brain injury have shown significant reductions in parenting stress, depression, and anxiety symptoms. However, in-person delivery of ACT poses barriers, particularly for parents of children with special needs. The COVID-19 pandemic has highlighted the need for virtual healthcare options, and smartphone-based ACT platforms provide a viable solution. Utilising smartphones allows convenient access to ACT interventions, accommodating parental caregiving responsibilities, overcoming traditional treatment limitations, and reducing resource requirements for healthcare systems. Studies indicate parents' interest and willingness to incorporate technology into their care, with smartphone-based ACT interventions demonstrating positive effects on parental well-being, psychological flexibility, mindfulness skills, and children's quality of life. Yet, challenges persist in maintaining user engagement in these publicly accessible mental health initiatives, with over 70% of individuals discontinuing use shortly after downloading. Pai.ACT To address these gaps, the current study proposes a pilot investigation of "Pai.ACT," a smartphone-based application designed to engage more than 1,000 Hong Kong-based parents. Funded by the Information and Technology Fund (ITF), Pai.ACT employs a conversational agent using voice-to-text capabilities to simulate ACT counsellors' responses in Cantonese. This application serves as a comprehensive digital ACT platform, offering multiple tiers of intervention, from self-help materials to personalized video conferencing sessions grounded in ACT principles. The platform's development incorporated analysis of an extensive dataset of over 10,000 tagged text instances from individual ACT sessions to train a deep-learning language model, facilitating precise psychotherapeutic interactions tailored to parents' needs. STUDY AIMS AND HYPOTHESIS TO BE TESTED The aim of this two-arm, repeated-measure, randomised controlled trial is to assess the feasibility, acceptability, and potential efficacy of a deep-learning, smartphone-enabled Acceptance and Commitment Therapy (ACT), referred to as Pai.ACT, is specifically designed for parents of children with special needs. This study seeks to determine whether this technological intervention can foster enhanced mental well-being for these parents over a three-month post-intervention period. This study hypothesizes that the implementation of Pai.ACT will lead to substantial improvements in mental well-being and psychological flexibility for parents of children with special needs. Compared to the control group receiving treatment-as-usual (TAU), we anticipate the intervention group receiving Pai.ACT will significantly ameliorate their mental health status immediately after the intervention and at a three-month post-intervention follow-up. This will underscore the potential of Pai.ACT as an effective and accessible tool for psychological support and resilience-building in this population.

Autism Spectrum Disorder, Attention Deficit Hyperactivity Disorder, Neurodevelopmental Disorders, Intellectual Disability, Specific Learning Disorder, Communication Disorders, Motor Disorders

Deep learning, Smartphone-delivered Therapy, Acceptance and Commitment Therapy, Parents, Children with Special Needs, Feasibility trial, Parental Support, Artificial Intelligence in Healthcare, Psychological Interventions, Parenting Stress

The randomisation procedure will be implemented using sequentially numbered, opaque, and sealed envelopes containing number cards (1=intervention, 2=control). This procedure will be carried out by a clerical staff member who has no involvement in the research activities, using a set of random numbers generated by a statistician. The randomisation process will remain concealed from the research team. Research Assistant (RA), blinded to the subject selection, will open the envelopes only after informed consent is obtained. Parents who have been randomised will be sent a secure link via email to assess the materials based on their study groups. All outcomes will be assessed through self-reports using the online tools embedded within the Pai.ACT application.

Parents allocated to this comparator group will receive conventional familial support offered by the hospital's Children with Complexity Community Support Programme (CCCSP) and allied Non-Governmental Organizations (NGOs). This support encompasses disseminating educational content focused on the management of children's affective and behavioral manifestations.

The Pai.ACT mobile app is an innovative therapeutic tool that utilizes Acceptance and Commitment Therapy (ACT). It integrates a sophisticated algorithm to analyze self-reported data and conversation texts when the user interacts with the AI chatbot and identifies what psychological inflexibility processes are required to be the most essential to be addressed for process-matched ACT interventions. These interventions, including self-help modules and experiential exercises, are enhanced with dynamic animations and audio metaphors. The app also offers 4-6 video-based individual ACT sessions with a dedicated counseling team and a referral system for intensive psychological support, all underpinned by empirical evidence from a previous ACT trial in Hong Kong.

Both the Pai.ACT group and the control group will receive conventional familial support offered by the hospital's Children with Complexity Community Support Programme (CCCSP) and allied Non-Governmental Organizations (NGOs). This support encompasses disseminating educational content focused on the management of children's affective and behavioral manifestations.

Parenting stress will be assessed using the Parental Stress Scale (PSS, 18-item, 5-point scale). The PSS is designed to evaluate the perception of parenting roles in terms of parental satisfaction and strain. A higher score represents a higher level of parental stress. The Chinese PSS version, including its subscales, has demonstrated strong convergent validity, discriminant validity, and internal consistency (Cronbach's alpha = .79-.88) among Chinese parents.

The Strengths and Difficulties Questionnaire (SDQ) will assess the child's emotional and behavioural symptoms. The SDQ uses a 3-point Likert scale to assess the child's emotional symptoms, conduct problems, hyperactivity/inattention, peer problems, and prosocial behaviour, with five items per subscale. The Chinese SDQ version has reported adequate test-retest reliabilities (Intraclass Correlation Coefficient = .75-.86) and discriminant validity among Hong Kong parents.

Anxiety symptoms will be assessed using the Generalized Anxiety Disorder-7 (GAD-7) scale. The GAD-7 is a 7-item, 4-point Likert scale with a total score of ≥10, indicating potential anxiety disorder.

Depressive symptoms will be assessed using the Patient Health Questionnaire (PHQ-9). The PHQ-9 is a 9-item, 4-point Likert scale with a total score of ≥10, indicating potential depression.

Psychological flexibility will be assessed using the Psyflex questionnaire and/or Comprehensive Assessment of Acceptance and Commitment Therapy (CompACT). The Psyflex is a 6-item, 5-point Likert scale designed to measure psychological flexibility, while the CompACT measures the core processes of ACT, including openness to experience, behavioural awareness, and valued action. Each item is rated on a 7-point Likert scale, with higher scores indicating greater psychological flexibility.

Perceived usability of the mobile app will be measured using the System Usability Scale (SUS). The SUS is a 10-item, 10-point Likert scale. The total SUS score ranges from 0 to 100. A score of at least 80 out of 100 indicates high perceived usability. This score would suggest that the mobile app, integrated with the 'Pai.ACT' mental health advisory system, is highly usable for the end users. These end users are primarily parents of children with Special Health Care Needs (SHCN).

Inclusion Criteria: Primary caregivers who are Cantonese-speaking Hong Kong residents. The caregiver must cohabitate with the child with special needs. The child under the participant's care should be aged between 2-8 years. The child should be either diagnosed or suspected to have one of the developmental conditions such as ASD, ADHD, or DD. These conditions must be recognised by the Child Assessment Service of the Department of Health and conform to the DSM-5 criteria. The diagnosis or suspected diagnosis should be documented in the electronic medical record at the study hospital or the case profile record at the collaborating non-governmental organisations. Note: The age range of 2-8 years was selected due to the substantial impact of parenting on the developmental milestones of preschoolers and junior school-aged children. Exclusion Criteria: Parents diagnosed with severe mental illnesses are excluded. Parents who are currently pregnant are excluded. Parents who are less than six months postpartum are excluded. Parents with a developmental disability that interferes with their ability to comprehend the program's content are excluded. Parents with cognitive, language, communication, visual, or hearing impairments or disorders that could impede their understanding of the intervention content are excluded. Parents currently participating in other psychosocial, psychoeducational, or parenting interventions are excluded.