Table 1 Dual-track role-play codesign
From: A community-codesigned LLM-powered chatbot for primary care: a randomized controlled trial
Codesign | Tracks | Descriptions | Role-play activities and objectives |
|---|---|---|---|
Contextual understanding | Community | Identify and prioritize unmet needs, sociocultural barriers and infrastructural constraints in existing digital health tools. | Assume researcher roles to codify key challenges and delineate culturally relevant success metrics. |
Researcher | Map technical capabilities/constraints of LLMs and establish ethical/regulatory compliance requirements. | Adopt user personas to simulate real-world interactions with existing chatbots, documenting usability gaps. | |
Cocreation | Community | Codevelop interaction modules with researchers, leveraging community expertise. | Lead prototype ideation workshops, prioritizing accessibility for low-literacy populations. |
Researcher | Translate community-derived requirements into technical specifications. | Conduct ethical impact assessments to mitigate risks. | |
Testing and refinement | Community | Provide qualitative feedback on interaction experiences with developed prototypes. | Codefine patient-centred evaluation metrics. |
Researcher | Iteratively refine prototypes, using community-informed simulation testing and performance audits to optimize functionality and usability. | Perform heuristic evaluations to proactively identify and address technical and experiential limitations. | |
Implementation and evolution | Community | Execute community-led pilot deployments to identify context-specific implementation barriers. | Analyse usage patterns to iteratively refine feedback instruments and optimize implementation protocols. |
Researcher | Conduct real-time equity audits of LLM outputs to address demographic performance gaps. | Stress-test system resilience via simulated high-load scenarios in resource-constrained environments. |
