Plurality Paradigm
The plurality paradigm emerged from the argument of 'How AI Fails Us' and has been developed institutionally through Allen's GETTING-Plurality network, Weyl's RadicalxChange foundation, and Tang's work on digital democracy in Taiwan. The paradigm is built on three structural commitments. First, intelligence is plural rather than singular—emerging from the interaction of diverse participants rather than from the centralized optimization of a single system. Second, technology should augment human cooperation rather than replace human cognitive labor. Third, the development and governance of AI should distribute authority broadly rather than concentrate it in a small number of organizations.
In the AI Story
The plurality paradigm stands in direct opposition to the dominant paradigm of contemporary AI development, which Allen's coauthors have called 'actually existing AI.' The dominant paradigm is characterized by massive centralization: a small number of corporations deploy enormous computational resources to train ever-larger models that are optimized to replace human cognitive labor across an expanding range of domains. The plurality paradigm challenges each of these features.
On centralization, the paradigm advocates for distributed development, open-source models, and governance structures that give affected communities authority over how AI operates in their contexts. This does not mean rejecting large models—some problems genuinely require scale—but it means refusing the assumption that only a few organizations can legitimately build and govern them.
On replacement, the paradigm advocates for AI designed to augment human cooperation and collective intelligence. This is visible in Taiwan's vTaiwan platform and related digital democracy tools, which use AI to facilitate large-scale deliberation rather than to automate decision-making. It is visible in Allen and Weyl's proposals for AI-enhanced public opinion research that makes hidden points of agreement visible to democratic publics.
On authority, the paradigm advocates for institutional structures that give diverse communities genuine governance voice. This includes data trusts that hold training data on behalf of the communities that produced it, multi-stakeholder governance of platform policies, and public investment in AI infrastructure as an alternative to monopolistic private control. The paradigm's institutional imagination is richer than either pure market competition or centralized state regulation.
The paradigm has gained momentum through the 2024 publication of Plurality: The Future of Collaborative Technology and Democracy by Weyl, Tang, and collaborators, and through the institutional work of GETTING-Plurality and related networks. It remains a minority paradigm in AI development, but its influence on policy discussions has grown significantly since 2021.
Origin
The plurality paradigm was articulated in 'How AI Fails Us' (2021) and has been developed through subsequent work by Allen, Weyl, Tang, and collaborators. The paradigm draws on the tradition of participatory design, commons governance, and democratic technology that Allen identifies as underlying 'many celebrated digital technologies such as personal computers and the internet.'
Key Ideas
Intelligence is plural. Genuine intelligence emerges from interaction among diverse participants, not from centralized optimization.
Augmentation, not replacement. Technology should expand human cooperation and collective intelligence rather than replace human judgment.
Distributed authority. AI development and governance should distribute power broadly across diverse communities and institutions.
Institutional imagination. The paradigm requires new institutions—data trusts, multi-stakeholder governance, public AI infrastructure—beyond the state-market binary.
Practical demonstration. Digital democracy platforms like vTaiwan demonstrate the paradigm's viability in practice.
Appears in the Orange Pill Cycle
Further reading
- E. Glen Weyl, Audrey Tang et al., Plurality: The Future of Collaborative Technology and Democracy (2024)
- Daron Acemoglu, Danielle Allen, Kate Crawford, E. Glen Weyl, 'How AI Fails Us' (Harvard Carr Center, 2021)
- Audrey Tang, Plurality (Stanford Digital Economy Lab, ongoing)