He takes social scientist Scott Page’s diversity prediction theorem (collective error = average individual error – prediction diversity) a step further to identify the three conditions which must be in place to know when crowds will predict well: diversity, aggregation and incentives:
Each condition clicks into the equation. Diversity reduces the collective error. Aggregation assures that the market considers everyone’s information. Incentives help reduce individual errors by encouraging people to participate only when they think they have an insight.
If you think about it, this is one of the core ways we derive value from the web. Consider:
The most interesting social areas of the web (especially but not exclusively horizontal networks) have diverse userbases.
Popular platforms aggregate by design – or if they don’t – external parties are building technologies on top of them