Complementarity Thesis

In the AI Story

The thesis was demonstrated empirically in Fung's studies of Chicago community policing and Porto Alegre participatory budgeting, where participatory processes produced outcomes — reduced crime, better-targeted infrastructure investment — that expert-only processes had failed to achieve. The mechanism was consistent across cases: practical knowledge about specific local conditions, available only to residents, proved essential to identifying problems accurately and developing responses that actually worked.

The thesis has direct implications for AI governance. The WeBuildAI project demonstrated complementarity in the algorithmic governance context: community stakeholders deliberating on algorithmic policy produced outcomes reflecting considerations expert-only governance had systematically neglected. The customer service representative's knowledge of how AI bias manifests in practice, the teacher's observation of students' engagement with AI writing tools, the parent's daily decisions about children's relationship with technology — all constitute practical knowledge that AI governance needs and that expert analysis cannot replicate.

The thesis addresses the expert governance objection: that participation trades off against decision quality in technically complex domains. The evidence shows the opposite — that in technically complex domains, expert governance without practical knowledge produces systematic failure modes that participation can address. The objection rests on a false dichotomy between expertise and participation; the thesis specifies their complementary relationship.

The thesis also explains why specific forms of practical knowledge matter differently in different governance contexts. AI in healthcare requires practical knowledge from patients and healthcare workers that differs from the practical knowledge relevant to AI in finance or AI in education. The Sectoral AI Governance Board design addresses this variability by creating domain-specific participatory mechanisms that access the specific practical knowledge relevant to each domain.

Origin

The thesis emerged from Fung's empirical observation that participatory governance succeeded where expert-only governance failed, combined with analysis of why this pattern occurred. The identification of complementary knowledge forms as the mechanism required integrating political-science analysis with philosophy of knowledge (particularly Michael Polanyi's work on tacit knowledge) and organizational theory (particularly work on distributed expertise in complex systems).

The thesis connects to broader work in participatory action research, citizen science, and co-production — all of which develop versions of the claim that expert and practical knowledge jointly produce outcomes neither can achieve alone. Fung's contribution was formalizing the claim as an empirically testable thesis with specific institutional implications.

Key Ideas

Two distinct forms of knowledge. Technical expertise and practical knowledge are genuinely different; neither reduces to the other.

Complementarity, not trade-off. The forms produce jointly superior outcomes; participation does not trade off against decision quality but enhances it.

Mechanism explains empirical pattern. The thesis specifies why participatory governance succeeds where expert-only governance fails, providing theoretical foundation for observed outcomes.

Domain-specific practical knowledge. Different governance contexts require different forms of practical knowledge, with implications for institutional design.