Power to the People? Opportunities and Challenges for Participatory AI (2209.07572v1)

Abstract: Participatory approaches to AI and ML are gaining momentum: the increased attention comes partly with the view that participation opens the gateway to an inclusive, equitable, robust, responsible and trustworthy AI.Among other benefits, participatory approaches are essential to understanding and adequately representing the needs, desires and perspectives of historically marginalized communities. However, there currently exists lack of clarity on what meaningful participation entails and what it is expected to do. In this paper we first review participatory approaches as situated in historical contexts as well as participatory methods and practices within the AI and ML pipeline. We then introduce three case studies in participatory AI.Participation holds the potential for beneficial, emancipatory and empowering technology design, development and deployment while also being at risk for concerns such as cooptation and conflation with other activities. We lay out these limitations and concerns and argue that as participatory AI/ML becomes in vogue, a contextual and nuanced understanding of the term as well as consideration of who the primary beneficiaries of participatory activities ought to be constitute crucial factors to realizing the benefits and opportunities that participation brings.

Opportunities and Challenges for Participatory AI

The paper "Power to the People? Opportunities and Challenges for Participatory AI" addresses the growing emphasis on participatory approaches within the field of AI and ML. It critically examines the implications of participatory methods in creating equitable, inclusive, and robust AI systems, especially in relation to historically marginalized communities. The authors—Birhane et al.—provide a nuanced exploration of participatory AI through historical context, case studies, and thematic discussion, emphasizing both its potential benefits and inherent challenges.

Historical Context and Emergence

Participatory AI is portrayed as a response to documented harms from AI systems in critical sectors such as security, justice, and healthcare. From a historical lens, the paper connects current participatory approaches in AI to broader participatory movements in technological development dating back to the 1970s Scandinavian workplaces. Participation historically aimed at workplace democracy, ensuring that the workers could have influence in technological adoption decisions. However, the paper acknowledges a persistent risk—participation may serve as a veneer without redistributing power effectively, drawing a cautionary analogy from colonial tactics where participation was co-opted to legitimize control. The authors stress that this historical backdrop is instructive for current participatory efforts in AI to avoid superficial engagement that perpetuates existing power imbalances.

Modes of Participatory AI

Through the examination of three case studies, the paper categorizes participation into three distinct typologies:

1. Participation for Algorithmic Improvement: In this mode, participation aims to enhance performance, quality, or diversity of technical components within AI systems—e.g., data labeling or cleaning.
2. Participation for Methodological Innovation: This involves leveraging participant-driven insights to inform or refine the design process, aiming more broadly at methodological improvements rather than discrete technical goals.
3. Participation for Collective Exploration: This form prioritizes participation as an end goal, fostering an emergent and iterative process that incorporates community needs, goals, and futures into decision-making.

These models illustrate how AI initiatives can vary in terms of engagement and impact, ranging from piecemeal data improvements to radical shifts in power dynamics.

Case Studies

The presented case studies serve as practical demonstrations of each mode:

• Machine Translation for African Languages: This grassroots initiative exemplifies participation that maintains community ownership and empowerment, addressing technical inclusivity for under-resourced languages.
• Māori Data Sovereignty: Reflecting high community control, this case underscores the importance of reciprocity and indigenous data rights in participatory processes.
• Dataset Documentation: Highlighting participatory methods for improving AI transparency and accountability, this case explores how participatory practices can refine technical accuracy and robustness.

Challenges and Considerations

The authors highlight multiple challenges in implementing participatory AI. These include managing power imbalances between stakeholders, avoiding cooptation where corporate interests may overshadow community benefits, and ensuring participation isn't merely tokenistic or perceived as "participation-washing." Furthermore, the paper acknowledges that participation must coexist with democratic governance, clearly delineating participatory mechanisms and democratic mandates.

Implications and Future Directions

Theoretical and practical implications of participatory AI are profound. As participatory approaches become mainstream, researchers must diligently differentiate participation from related constructs such as inclusion or mere consultation to avoid conflation and dilution of purpose. Moreover, there's an acknowledged need for developing metrics that effectively evaluate participatory efforts beyond traditional cost-benefit analyses.

Future developments in AI should stress continuous reflexive assessment, prioritizing vibrant participation across stakeholders to foster systems that indeed empower and serve the public equitably. Participation in AI presents an opportunity to align technological advancements with societal needs—particularly those of the marginalized—while challenging systemic inequities within AI development.

The paper calls for ongoing discourse and critical examination of participatory methods within AI, ensuring they genuinely enhance prosperity, justice, and shared ownership among communities they impact.