Frameworks, Methods and Shared Tasks: Connecting Participatory AI to Trustworthy AI Through a Systematic Review of Global Projects

As the discourse on responsible and trustworthy AI intensifies, Participatory AI (PAI) presents a compelling approach to the democratic development of automated technologies. But how should the field think about how and whether participatory methods increase trust in, and the trustworthiness of, AI systems? In response to the recent growth in PAI research, we conducted a systematic examination to understand the landscape of methods and theoretical lenses used in participatory AI projects, and to connect those methods and lenses to trust building. This paper analyzes 95 global PAI projects to understand the participatory landscape of trustworthy AI. Our review explores differences in theoretical frameworks, participation methods, and the details of shared tasks within the AI lifecycle across sectors and geographies. Our findings reveal an evolving definition of PAI, with actors implementing diverse methods and shared tasks. Focusing on shared tasks provides a lens for analyzing how participation can build trust in, and trustworthiness of, AI systems. Considering PAI through the lens of trust creates a framework for mapping PAI activities to both participant and broader public trust and illustrates gaps where PAI processes alone will not be enough to ensure trustworthy AI systems.