Lay Stakeholder Centric Sociotechnical Mechanisms for Addressing the Impacts of Generative AI

This work argues that current approaches to mitigating the harms of generative AI (genAI) overlook the perspectives of lay stakeholders: people without professional or technical expertise relating to genAI, but who possess a contextual and situational experience about how the impacts of these systems affect them on the ground level. While expert-driven frameworks dominate risk assessment, this research introduces a participatory, sociotechnical framework that redistributes agency to lay stakeholders through three mechanisms addressing different scales of harm, each presented through a different case study. At the micro-scale, a technological mechanism empowers users of generative music tools to resist harms resulting from uninformed attribution, such as copyright infringement. At the macro-scale, a policy mechanism integrates lay stakeholders' perspectives into the policy pipeline to inform governance of AI in relation to media ecosystems. At the meso-scale, a judicial mechanism strengthens courts’ ability to evaluate genAI-related copyright disputes by incorporating lay perspectives into evidentiary processes. Together, these mechanisms propose actionable ways to balance relationships between technology developers, governing bodies, and lay stakeholders, ensuring that those most affected by genAI are meaningfully included in addressing the harms.