Why Observed Public Stories Can Beat Direct Polling Questions
Hunt argues that polls ask people to predict or describe themselves, tasks at which respondents are often unreliable and sensitive to wording. Semantic analysis instead observes the meaning of what people are already discussing across local and national media, reducing dependence on a pollster's phrasing.
- People are poor at stating what they think and forecasting what they will do.
- Prediction markets improve incentives but can still be manipulated.
- Poll results can vary materially with question wording.
- Semantic signatures examine expressed meaning without asking a direct question.
“When we know people are terrible at both of them terrible”
“We're not doing polling. We're seeing what people are actually the meaning of what they're talking about in media.”