Deliberative polling reveals the public would hold different views if it thought harder

Evidence

James Fishkin (1991; 1995) argues that the media should at least occasionally attempt to conduct what he calls a deliberative opinion poll. This would involve taking a random sample of people, bringing its members together for some days in order to deliberate about the issues in question, broadcasting some of their debates and then polling participants at the end. The evidence is that people’s opinions change in the course of such deliberation, and it would surely be preferable if public opinion were represented, and shaped, by the evidence of deliberative polls, not just by the evidence of ordinary polling. ‘An ordinary poll models what the public thinks, given how little it thinks, how little it knows, how little it pays attention. A deliberative opinion poll, by contrast, models what the public would think, if it had a better chance to think about the questions at issue’ (Fishkin 1995: 45). — Philip Pettit, Republicanism, 767

Categorization

• QUESTION: Fishkin claims people’s opinions change under deliberation — worth chasing the actual empirical evidence base for that and how stable the changes are.
• SEED: Article angle: what a deliberative poll of an actual contested issue would reveal, and the case that ordinary polling manufactures public opinion rather than measuring it.
• XREF: connects to: Voter preferences respond to framing even when expected values are equal; Factual hearings rarely shift entrenched public opinion — Dorothy A. Brown, Getting to Reparations: How Building a…, 1145; Public opinion resists change while conversation invites being changed — Jenny Odell, How to Do Nothing: Resisting the…, 482 (+1 more)

Broader Connections

• Public opinion resists change while conversation invites being changed #corroborates (0.80) — Jenny Odell, How to Do Nothing, 482
• Factual hearings rarely shift entrenched public opinion #complicates (0.78) — Dorothy A. Brown, Getting to Reparations, 1145
• Voter preferences respond to framing even when expected values are equal #corroborates (0.72)