Deliberative Polling

In the AI Story

The methodology's distinctive contribution was combining representative sampling with intensive deliberation in a single empirical instrument. Previous polling measured pre-formed opinions without improving them; previous deliberative experiments often involved unrepresentative participants. Fishkin's design measured the specific transformation that informed deliberation produced in a representative population, providing the evidence base for claims that had previously rested primarily on theoretical argument.

The results across applications are consistent. Participants shift from positions shaped by immediate reaction toward positions reflecting considered engagement with complexity. Extreme positions tend to moderate; qualified positions tend to develop specificity. The shifts are not toward predetermined conclusions — different deliberative populations reach different conclusions on different issues — but toward the characteristics of considered rather than unconsidered judgment.

Applications to AI governance have been limited but illuminating. The WeBuildAI project and related initiatives have adapted deliberative polling methodology to algorithmic governance questions, demonstrating that non-specialists can engage meaningfully with technically complex questions when given adequate information and structured deliberative opportunities. The outcomes reflect considerations that expert-only governance had systematically neglected.

The methodology's significance for Fung's framework is foundational. The empirical claims about what the silent middle would produce if given institutional infrastructure rest on the evidence from deliberative polling. The methodology demonstrates that the ambivalence, nuance, and trade-off awareness that characterize the silent middle's individual judgments can be accessed and amplified through properly designed deliberative processes.

Origin

Fishkin developed the methodology at Stanford's Center for Deliberative Democracy beginning in the late 1980s. The first major national application was the 1994 British deliberative poll on crime; the 1996 American poll on the national issues convention at the University of Texas established the format in its current form. Subsequent applications have been conducted in dozens of countries on issues ranging from energy policy to constitutional reform.

The methodology's development drew on earlier work in democratic theory (particularly Dahl and Habermas) and empirical political science (particularly on the problem of uninformed public opinion). Fishkin's contribution was integrating theoretical insight with empirical methodology in a form that could produce replicable evidence about deliberation's effects.

Key Ideas

Random sampling plus deliberation. The combination measures the transformation informed deliberation produces in a representative population.

Deliberation produces predictable shifts. Participants develop nuance, trade-off awareness, and qualified positions regardless of the substantive topic.

Non-specialists engage meaningfully with complexity. Given adequate information and structured time, participants reason productively about technically complex questions.

The empirical foundation of deliberative theory. The methodology transformed deliberative democracy from theoretical aspiration to measured institutional achievement.