Understanding Professionals’ Needs in Integrating AI Chatbots for Wheelchair Assessment Services in Low- and Middle-Income Countries

Assistive Technology (AT), defined by the World Health Organization (WHO), is “an umbrella term for assistive products (e.g., wheelchairs) and their related systems and services” [98]. These services include assessment, fitting, training, follow-up, maintenance, and more, which are all essential to ensuring the safe and effective use of ATs like wheelchairs. However, in low- and middle-income countries (LMICs), wheelchair service provision is often hindered by limited access to trained professionals, inadequate supply chains, and insufficient infrastructure [30, 99]. To address these challenges, the WHO has established guidelines for wheelchair provision [97] and training packages tailored for low-resource settings [105], aiming to improve service capacity and quality.

 

Meanwhile, growing research interest in digitalizing healthcare services, such as developing electronic health records (EHR) systems [14, 65, 69] and telemedicine, [24, 52, 64, 82] shows how digital health tools could improve healthcare practice. Amid this rising enthusiasm [81], the “Chatbot Tsunami” [33] has arrived, with Artificial Intelligence (AI) chatbots riding the crest of the popularity wave in healthcare research, especially after the release of GPT-3 in June 2020. Numerous studies have highlighted the promise of AI chatbots in supporting healthcare professionals to improve patient care [57, 59], enhancing professional workflows by offering insights to aid diagnoses and offload administrative tasks [41, 70, 75, 85].

 

However, limited attention, especially in Human-Computer Interaction (HCI) and AT, has been paid to investigating how such tools might support wheelchair provision in LMICs. Mo et al. [58] offered an early exploration, indicating the potential of large-language model chatbots in improving wheelchair service by providing physical therapists with real-time consultation and training. While promising, it remains unclear how such chatbots should be designed to realize these opportunities in practice. For instance, what would be the practice-specific challenges that affect the integration of chatbots into existing workflows? What kinds of chatbot interactions would amplify rather than disrupt professionals’ expertise? Thus, there is a need to better understand how to transform the promises of AI chatbots into workflow-compatible and context-aligned tools for wheelchair provision services.

 

To explore these questions, we conducted a two-part study, shifting from broad exploration to a more specific, practice-based approach, focusing on wheelchair assessment services with 11 physical and occupational therapists from Africa and South Asia. Wheelchair assessment plays a central role in ensuring appropriate provision [105]. First, we interviewed professionals to understand the challenges in current practices. Second, we encouraged participants to reflect on the imagined use of two chatbot design probes: "Ask Wheelie" for supporting wheelchair assessments and "Wheel Care" for facilitating personalized follow-up care. Based on the feedback, we present six main themes that reveal professionals’ core need for digitalization and highlight varying tensions they anticipated in introducing AI chatbots into their workflow, such as concerns over losing control of practice workflows, the difficulty of conveying contextual nuance through chatbot interfaces, and the risk of disrupting trust and rapport with clients. Based on our findings, we identified 13 tensions and categorized them into three interconnected domains: Practice Structure, Anchored Values, and Contextual Readiness. We propose them as a tension-informed design framework and discuss the importance of building human capacity when designing digital solutions for AT service in resource-constrained settings.