The Wisdom of Crowds

Context:

‘Why the Many are smarter than the Few.’  James Sorowiecki [11]

•••

Where the average chance of a member of a voting group making a correct decision is greater than fifty percent the chance of the group as a whole making the correct decision will increase with the addition of more members to the group.  The Condercet Jury Theorem

In his best selling book, ‘The Wisdom of Crowds’, James Sorowiecki explains a simple yet counter-intuitive idea that has profound implications - that aggregating information from a large number of sources is likely to yield a better decision than asking an expert or a small group of experts. He writes: ‘With most things, the average is mediocrity.  With decision making, it's often excellence.  You could say it's as if we've been programmed to be collectively smart.’ [11]  Sorowiecki provides examples from price markets, TV game shows and other anecdotes to show how this process works, and to reveal its limitations so that we understand when it won’t work.  According to Sorowiecki, good results are likely when the group has sufficiently diverse opinions so that errors will tend to cancel out, members of the group are independent and not swayed by opinion leaders, there is no centralising hierarchy, and where a mechanism exists for aggregating individual judgements into a collective decision.

The fact that in certain conditions groups can be smarter that even the smartest people in them can be proved using a simple statistical formula, which leads to the Condercet Jury Theorem.  For example, in a group of three people, where the average person has a 67 per cent probability of having the correct answer, the probability of the group being right is 74 per cent; if the group is larger than three and the individual average probability remains, the probability of the group having the correct answer is even higher than 74 per cent, and will eventually approach 100 per cent.  But the Theorem also works the other way – when the average individual probability of being correct is less than 50 per cent, the cumulative probability of the group being correct tends to zero as more misinformed people join the group. 

The Condercet Jury Theorem applies strictly to statistical voting groups, where the decisions of individuals in the group are independent of the rest of the group, and where the decision is a simple choice; for example, deciding whether or not a person is guilty or guessing the weight of a cake.  However, the judgement of statistical groups can be poor if there is a systemic bias in the group towards a particular choice, or if the individuals in the group lack the necessary information and can only make a random choice.  In such cases the group may not perform as well as an expert.  The question then is whether deliberative groups (i.e. groups which debate together) can regularly outperform individual experts. 

In a deliberative group, the aggregation of information is a more complex process than the situation described by Sorowiecki or the Condercet Jury Theorem because it is shaped by social pressures.  Deliberative groups can fail spectacularly when these forces lead to polarised positions or ‘groupthink’ – the title of Surowiecki’s book is actually a play on an earlier history by Charles Mackay: ‘Extraordinary Popular Delusions and the Madness of Crowds’.  But a positive aspect of deliberative groups is the ability of their members to learn from each other.  According to Cass Sunstein: ‘The most general point in favour of deliberation is that a deliberating group will converge on the truth … if the truth has some initial social support within the group and when the task has a demonstrably correct answer according to a framework that group members share.’ [12]

Therefore:

The issues within this pattern fall into two categories: simple aggregation of information (as in a voting system) and the more social context of collaborative working: when writing collaborative documents on the wik, provide a for deliberative debate attached to the text; consider the use of a scoring mechanism for collectively rating articles published on the wiki – e.g. Amazon’s system of customer reviews.

•••

The pattern Peer Production relies upon the The Wisdom of Crowds, but working against this relationship is the pattern Unequal Participation.