Workers Participating in Transparency: Addressing the Gap in AI Transparency on Labor

Transparency is one of the few practices that AI research and industry agree upon and consistently use to advance safe and responsible AI. Yet despite the visible societal attention to the economic impacts of AI, efforts on transparency largely omit impacts on workers. To address this gap, we must include workers as a topic of transparency, as an audience for transparency, and as participants in transparency. We do so by applying concepts of human-centered transparency and transparency as a process, while integrating insights from participatory AI to emphasize transparency as a participatory process. Using examples from the field, we illustrate how to center transparency on human understanding and the needs of different stakeholders, addressing workers as an audience and topic of transparency. We show examples of transparency as a participatory process, not just a product, integrating workers and their domain expertise throughout the full machine learning lifecycle of design, development, and deployment. In particular, union collective bargaining around technology offers a model for participatory processes that strengthen transparency and empower workers. The more workers are empowered in transparency as a process, the more transparency can go beyond a checklist of produced artifacts and become a process for shared governance.