Real laziness here, stealing notes from a review of a book that I can’t be bothered to read
The Book: The Wisdom of Crowds: Why the Many are Smarter than the Few by James Surowiecki
The Review: London Review of Books. The Notes: gyford.com
Phenomenon: For many decisions the average of many judgements is often better than the judgement of a single, albeit expert, individual. Example: judging the number of jellybeans in a jar – typically even individuals who have previously been most accurate (‘the experts’) will be outperformed by the average answers.
Seems analogous to the ‘less is more’ effect. This is, roughly, that sometimes an overabundance of information can distract you from applying an on-average-correct heuristic. Example: Answer this question Which city has more inhabitants: San Diego or San Antonio?. Who should do better at this question, Germans or Americans? The Germans, typically, have little knowledge of the size of American cities. So when given this pair they guess that the one they have heard of is larger (San Diego), and they are correct. The Americans know lots about American cities. They try to use all the information they have to make a correct decision. Which is more politically important? Which has more people I know living in it? Which felt bigger when I visited? Sometimes this information is helpful, sometimes it is distracting. Sometimes decision making based on more knowledge is outperformed by that based on less knowledge (here the analogy with the wisdom of crowds i guess). In one study  the German group using their simple recognition heuristic scored 100%. More generally, often neither method/group is always correct, but the simple, one criterion, rule can often be more correct.
So one mechanism by which the wisdom-of-crowds effect works is probably just reducing the level of knowledge that is contributing to the decision. A dumb kind of wisdom!
But crowds can often be dumb-dumb too, especially when they become herds. What are the conditions under which they keep their dumb-wisdom, the conditions when a mixture is better than the best expert?
Requires certain conditions for the crowd to make good decisions: members of group must be willing to think for themselves; they must be mostly independent of each other; must be reasonably decentralised; must be some means of aggregating opinions into a collective judgement. If people start second-guessing each other, or following each other, the crowd becomes a herd and herds are bad at decision making…Crowds do not do well the question is not a straight-forwardly cognitive one. They are not good at moral judgments.
You might argue that a group of people which are all thinking for themselves isn’t really a crowd. You might also argue that the wisdom of crowds doesn’t apply to moral judgements because individual judgements are non-commensurable in so many ways, not just because they are subject to lots of weird biases. If the choice is the same, but the individuals are making different decisions (e.g. they have access to contradictory information and/or they are using different criteria to select what a ‘good’ answer is) then aggregation isn’t possible.
I think a more helpful book would not be The Wisdom of Crowds – Why the Many are Smarter than the Few but The Wisdom of Crowds – How the Many can be Smarter than the Few. Anyway, good to have some starting notes on when crowd decisions will outperform individual decisions – and when ’emergent’ decisions will be herd-like and unproductive.
 Goldstein, D. G., & Gigerenzer, G. (2002). Models of ecological rationality: The recognition
heuristic. Psychological Review, 109, 75-90. Online here