Session 2

AI, Health, and Mental Health Social Work

Moderator: Dr. Yang Wang (Associate Professor, City University of Macau, China)

O2.1  Knowledge and stigma of HIV/AIDS among people living with mental disorders in China: Challenges for social workers

*Tianming Zhang¹, Yin-Ling Irene Wong²

¹Shanghai University, China; ²University of Pennsylvania, USA

Abstract

Background: People living with mental disorders (PLWMD) may face elevated vulnerability to HIV infection, yet their knowledge of HIV/AIDS and attitudes toward people living with HIV/AIDS (PLWHA) remain underexplored in rural China, especially in high-risk areas. At the same time, PLWMD themselves often experience mental illness stigma, which may intersect with HIV-related stigma and further shape their perceptions and attitudes. In order to shed light on the practice areas where Chinese social workers could provide service in educating and preventing HIV/AIDS as well as reducing stigma, the current study is to assess knowledge of HIV/AIDS among people living with serious mental disorders, to document perceived stigma of mental disorders and HIV-related stigma, and to identify factors that are associated with HIV-related stigma, including HIV/AIDS knowledge, and perceived stigma of mental illness.

Methods: This study used a cross-sectional survey design. A total of 147 individuals with serious mental disorders were recruited from a high-risk rural county in China. Data were collected using structured questionnaires measuring HIV/AIDS knowledge, perceived stigma related to mental disorders, and stigmatizing attitudes toward PLWHA. Descriptive statistics were used to summarize participants’ levels of knowledge and stigma. Regression analysis was conducted to examine the association between HIV/AIDS knowledge and attitudes toward PLWHA.

Results: Participants demonstrated generally low levels of HIV/AIDS knowledge, with correct response rates to knowledge items ranging from 14.3% to 66.7%, indicating a lower level of knowledge than that reported in other Chinese populations. Perceived stigma of mental illness and stigmatizing attitudes toward PLWHA were both common in this sample, and the level of negative attitudes toward PLWHA appeared higher than that reported in other populations in China. Regression results showed that participants with higher levels of HIV/AIDS knowledge were more likely to report more positive attitudes toward PLWHA.

Conclusions: The findings suggest that PLWMD in rural high-risk areas represent an underserved population in HIV/AIDS education and stigma reduction efforts. Limited HIV/AIDS knowledge and prevalent stigmatizing attitudes highlight the need for targeted interventions addressing both information gaps and prejudice. Social workers in China are well positioned to promote HIV/AIDS knowledge, raise awareness among PLWMD, and challenge HIV-related stigma through community-based education and anti-stigma practice. Future research may further explore the mechanisms linking knowledge and stigma and develop culturally responsive interventions for rural mental health populations.

O2.2  The Acceptance of Conversational Artificial Intelligence Among People with Dementia and Their Caregivers: A Systematic Review

*Keyi Li¹, Huanran Liu¹, Peiyi Lu¹

¹Department of Social Work and Social Administration, The University of Hong Kong, Hong Kong, China

 Abstract

Background and Objectives: The rising global prevalence of Mild Cognitive Impairment (MCI) and progressive dementia places significant strain on caregivers and healthcare infrastructures. Conversational Artificial Intelligence (AI), including Large Language Model (LLM)-powered and Natural Language Processing (NLP) applications, offers scalable solutions for personalized cognitive stimulation, daily assistance, and psychoeducational support. However, real-world integration remains limited, as development is often prioritized by engineering capabilities rather than user-centered needs. This systematic review aims to identify the primary facilitators and barriers influencing the acceptability, continuous engagement, and long-term adoption of Conversational AI among individuals with MCI or dementia, alongside their formal and informal caregivers.

Methods: This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A comprehensive literature search was executed across major academic databases to identify relevant peer-reviewed studies. The review included non-randomized and mixed methods designs evaluating digital or robotic systems equipped with conversational AI, such as generative chatbots, smart voice assistants, and socially assistive robots utilizing NLP or LLMs. Target populations comprised individuals formally diagnosed with MCI or dementia, and their caregivers, across domestic, community, and institutional care settings. Two independent reviewers managed study selection, data extraction, and methodological quality appraisal. Data concerning user acceptability, continuous engagement, and system usability were extracted and analyzed via narrative synthesis to elucidate cross-cutting themes regarding technology adoption.

Results: The synthesized literature reveals that Conversational AI serves as a powerful tool for providing emotional companionship, reducing loneliness, and facilitating diversion therapy. The integration of advanced LLMs has significantly enhanced conversational fluency, naturalness, and contextual relevance, which drives higher user engagement and acceptability. Conversely, primary barriers to adoption include significant usability challenges, such as rapid synthesized speech, poor speech recognition (especially for regional accents or impaired speech), and complex user interfaces. More specific conclusions require further research.

Conclusions and Implications: Conversational AI holds substantial promises for improving the quality of life for PwD and alleviating caregiver burden by offering accessible, scalable, 24/7 support. However, usability barriers and ethical concerns regarding safety and privacy currently impede sustained, long-world adoption. By bridging the gap between human-computer interaction and gerontology, this review highlights the urgent need for user-centered, culturally tailored designs and robust clinical validation. Future development must involve collaboration among engineers, clinicians, and social workers to ensure these AI applications are genuinely accessible, safe, and effective for the dementia demographic.

O2.3  Oneiro: An AI-Driven Multimodal Companion for Mitigating Peri-procedural Anxiety in Pediatric OSA Patients – A Pilot Study

*Jiumo Wang¹, Huiping Luo², Yue Zou¹, Hui Fa3

¹Shanghai Tech University, China; ²Fudan University, China; ³Shanghai Tech University, China; 3City University of Macau, China

Abstract

Introduction: The diagnosis and treatment of Pediatric Obstructive Sleep Apnea (OSA) are often complicated by peri-procedural anxiety, leading to non-compliance and "white coat syndrome." Based on the principle that familiarity fosters cooperation, we developed "Oneiro," an AI-enhanced therapeutic system grounded in theories of Embodied Interaction and Narrative Psychology. Crucially, the synergy between AI and design lies in "operationalizing" empathy: Generative AI (AIGC) serves not merely as a content producer but as a dynamic mediator that translates the child’s physical reality into a narrative form. By analyzing visual attributes to generate a "mirrored" avatar, it bridges the physical-digital divide to foster immediate trust. This study evaluates Oneiro's feasibility in utilizing generative narratives and tangible interactions to reframe the medical procedures, thereby improving patient adherence and reducing distress.

Materials and Methods: The Oneiro system, co-developed by a technical university and a specialized tertiary medical center, employs a multimodal framework combining AIGC and IoT. The intervention follows a three-phase embodied therapeutic model designed to address specific psychological needs:

Pre-Procedural Trust Establishment (Empathy & Belonging): Utilizing computer vision to analyze the child's visual attributes (e.g., clothing style and color), AIGC generates a personalized "mirrored" digital avatar. This visual alignment bridges the digital-physical divide, fostering immediate empathy and a sense of belonging.

Intra-Procedural Support (Companionship & Courage): The avatar provides real-time accompaniment via a mobile interface. By integrating narrative guidance with rhythmic auditory stimulation, the system offers constant companionship to alleviate isolation and foster courage during distressing procedures.

Tangible Reinforcement (Achievement & Adherence): A post-visit token economy rewards patients with NFC-enabled physical collectibles. This interaction fosters a sense of achievement and encourages long-term adherence by creating anticipation for collecting new digital assets during future follow-up visits.

A randomized pilot study was conducted with 13 pediatric OSA patients (aged 4–12). Patients were randomized to the Oneiro intervention group or a control group receiving standard nursing care. Anxiety levels were assessed using the modified Yale Preoperative Anxiety Scale (m-YPAS), and qualitative feedback was collected.

Results: The intervention group demonstrated a statistically significant reduction in anxiety scores on the modified Yale Preoperative Anxiety Scale (m-YPAS) compared to controls (Mean: 28 ± 5 vs. Control: 39 ± 7; p < 0.05). Qualitative: Patients exhibited enhanced narrative coherence, effectively reframing the medical procedures as shared adventures. This psychological shift was evidenced by spontaneous verbalizations, such as one patient noting, "My rabbit friend came with me," indicating successful emotional transference. Adherence: The tangible-digital feedback loop led to a significant improvement in motivation for follow-up visits and long-term engagement.

Conclusions and Implications: This preliminary pilot study demonstrates that emotionally resonant, multimodal systems like Oneiro can effectively improve the experience of pediatric patients. By transforming medical procedures into gamified narratives, Oneiro not only reduces peri-procedural anxiety but also provides a promising framework for long-term adherence management. This study highlights the potential of integrating AIGC and IoT to support comprehensive, human-centric care in pediatric chronic conditions.

O2.4  A Study on Occupational Health Risks of New Employment Groups and Their Social Work Intervention Pathways

* Jiachen Huang¹

¹School of Social Development, East China Normal University, China

Abstract

Background and Purpose: Against the backdrop of digital economy expansion, the scale of new employment groups, represented by food delivery riders, has grown rapidly. Their occupational health issues have become a critical social concern, yet empirical research and targeted social work intervention pathways remain insufficient. This study aims to explore the core influencing factors and formation mechanisms of occupational health risks among food delivery riders, and to construct a multi-level social work intervention framework, filling the empirical gap in occupational health research for new employment groups.

Methods: This mixed-methods study recruited 173 food delivery riders in Shanghai as research participants. Data were collected through questionnaire surveys and in-depth interviews. An integrated analytical framework was built based on ecosystem theory and resilience theory. Quantitative analyses, including correlation analysis, linear regression, and Lasso regression, were conducted to identify key risk factors, while qualitative thematic analysis was applied to interpret interview data for in-depth understanding of health dilemmas and underlying mechanisms.

Results: Night-shift frequency and work-related injury incidents were identified as the two core negative predictors of riders’ occupational health, with work-related injuries exerting a stronger adverse impact. These factors led to both explicit health problems (e.g., circadian rhythm disorders, heightened psychological anxiety) and implicit harms (e.g., sleep sensitivity, psychological trauma). Variables such as employment tenure and employment type showed differentiated effects across health dimensions. Fundamentally, occupational health dilemmas stemmed from imbalanced interactions across micro, meso, and macro systemic levels.

Conclusions and Implications: This study validates the applicability of ecosystem and resilience theories in occupational health research for the new employment sector. A three-tiered social work intervention pathway was developed: micro-level individual empowerment, meso-level platform collaboration, and macro-level policy support, accompanied by a collaborative implementation mechanism involving social workers, platforms, governments, and communities. The findings provide theoretical support and practical solutions for improving occupational health outcomes and refining social work service systems for new employment groups, contributing to inclusive and resilient social development in the Asia-Pacific context.

O2.5  Loneliness and Nomophobia Subtypes among Chinese College Students: A Latent Profile Analysis

*Juanjuan Wang¹

¹Hubei University of Arts and Science, China

Abstract

Background and Purpose: Nomophobia refers to the anxiety and fear experienced when individuals are unable to access or use their mobile devices, or when they are disconnected from them. Nomophobia is prevalent among college students, yet prior studies have primarily relied on total scores, which overlook the heterogeneity within this population and limit the utility of findings for targeted interventions. The present study aimed to identify latent subtypes of nomophobia among college students and to examine the predictive role of loneliness in higher-risk classifications.

Methods: Participants were 899 college students in China recruited through convenience sampling. Data were collected using self-administered questionnaires measuring nomophobia, loneliness, and covariates. Latent Profile Analysis (LPA) was conducted to identify subgroups of students with distinct nomophobia patterns. The optimal profile solution was determined based on model fit, classification quality, and interpretability. The R3STEP three-step multinomial logistic regression approach was then used to examine whether loneliness predicted latent profile membership while accounting for classification error, with covariates included in adjusted models.

Results: Results indicated a four-class solution as optimal: C1 (low-fear type), C2 (moderate-low fluctuating type), C3 (moderate-high fear type, the largest group, 39.8%), and C4 (high-fear type). After adjusting for covariates, loneliness was significantly and positively associated with membership in C2, C3, and C4, indicating that higher levels of loneliness were more prevalent among students classified into higher-risk nomophobia subtypes compared to the low-fear group.

Conclusions and Implications: The findings suggest that loneliness is a prominent psychosocial characteristic across higher-risk nomophobia subtypes and should be incorporated into early screening and intervention planning. For social work practice, the results support stratified and mechanism-oriented digital health strategies: loneliness screening and connection-building may serve as universal intervention modules, while additional supports can be tailored to subtype-specific needs.