A Sustainable Path for AI Development to Empower Communities and Serve Public Interests
Policy Paper
Feb. 18, 2025
RethinkAI, a multi-partner initiative that is part of New America’s Technology and Democracy programs, cosigns a new paper by Thomas Krendl Gilbert, CEO of Hortus AI and member of RethinkAI’s Civic AI Advisory Trust, that outlines a vision to empower local communities to integrate AI technologies on their own terms.
Generative artificial intelligence (AI) is rapidly expanding and becoming deeply embedded in everyday life. However, a growing disconnect between AI company decisions and public interests is eroding public trust. This decline is driven by the public’s loss of agency in shaping AI’s role in society. Current AI alignment methods, such as Reinforcement Learning from Human Feedback (RLHF), treat the public as a passive data source rather than an active participant. AI companies extract preferences to optimize models for their own technological and financial goals without meaningful public influence over AI’s purpose or governance. This approach is unsustainable, both in terms of public trust and long-term market viability.
But there is another way. We propose a new paradigm: Public and Responsible AI through Societal Empowerment (PRAISE). Instead of aligning humans to AI, PRAISE aligns AI development with publicly defined goals. This requires shifting decision-making power from corporations to the public, ensuring AI serves societal needs rather than private interests.
PRAISE presents a new division of responsibilities for AI development. It comprises an active, dynamic feedback loop between the public and AI vendors through four key steps:
- Distill: A public entity provides a specification for AI’s intended purposes;
- Disclose: The public stakeholder decides what must be disclosed about AI capabilities, risks, and trade-offs, while the AI vendor provides that information;
- Deliberate: The public decides on a path forward; and
- Deploy: The AI provider executes in ways that realize the public’s goals and needs.
Through PRAISE, the public is responsible for deciding how AI should work, and companies are accountable for ensuring it works as intended. These are not new ideas. PRAISE draws inspiration from earlier periods of rapid technological change in which stakeholders rose up to voice and defend the norms of public health and safety. This moment calls for rearticulating these values, not abandoning them for the sake of evermore powerful models.
PRAISE offers a sustainable path for AI development—one that fosters trust, aligns with societal values, and ensures long-term market success. This shift redefines AI alignment as a public-centered process rather than a corporate-driven optimization problem. Ultimately, PRAISE is a trellis on which new public values can grow. Hortus AI aims to build that trellis, providing a framework and tools to ensure that AI continually serves the needs of various public interests.