Up there with my P&G love affair and my love for markets is a love for the wisdom of crowds thesis.
However, I’m starting to find the thesis being used far too often in odd instances where it’s taken out of context and applied improperly.
That’s why I was happy when I saw an article in this month’s HBR titled When Crowds Aren’t Wise.
It turns out that some old French guy, Marquis de Condorce, is famous for the Condorcet jury theorem that shows that crowds trend towards correct decisions when each person is more likely than not to be correct:
“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.
This means that crowds aren’t wise (my friend Howard will like this) when “each individual in a group is more likely to be wrong than right because relatively few people in the group have access to accurate information.” As the size of this crowd grows the chance that they’ll forecast the correct outcome falls to zero.
Now to bring this back to another interest of mine: personal filters and the discovery of cultural consumables (movies, books, goods, …).
For discovery is it better to depend on experts or rely on the wisdom of crowds?
André makes a compelling argument that friends are wiser than crowds. (via Flying Seeds)
But think about the problem of cultural recommendations (shopping/music/…) wrt Condorcet’s theorem:
- It’s fair to assume that individuals within a crowd will have a less than 50 % chance of knowing what you will like. Therefore as the crowd increases the likelihood of a “correct” cultural recommendation falls to zero.
- Friends (ppl with shared cultural interests) have a > 50 % chance of knowing what you will like. The problem is that the low number of individuals with shared interests doesn’t drive constant high-accuracy recommendations.
What’s the solution? Personal filters - weighted coefficients for the wisdom of: trusted subject matter experts, friends, a larger group of “friends”, crowds.
Take my (rudimentary) personal filter for music discovery:
1) There are a few people, with a similar taste in music, who discover great new music that I love. Recommendations from these individuals carry a lot of weight.
2) I have friends who like discovering new music but don’t share identical tastes in music. While they recommend good music, additional filtering needs to occur.
3) Last.fm aggregates my listening habits, comparing them to others in the network to offer me algorithm-based recommendations. I’ve discovered some amazing bands through the service but I still filter out a lot of noise.
4) Sometimes I discover great new music through crowd based discovery methods. Far from perfect I generally filter out most of the recommendations.
It’s an imperfect process filled with inefficiencies but imagine a service that takes (1), (2), (3), and (4) and runs an algorithm such as: W(1) x X(2) x Y(3) x Z(4) to filter and offer recommendations.
With respect to the Condorcet theorem, personal filters aggregate large numbers of individuals who tend to be correct > 50 % of the time, applying a heavier weighting to those with a higher accuracy rate, to produce the right suggestion. In this way you leverage the value of both crowds and experts/friends to get accurate cultural recommendations.
[it's worth noting that the Condorcet theorem doesn't translate well to predictive markets where trading occurs: "traders are self-selected based on their degree of confidence in their beliefs, and the extent to which those beliefs differ from the market judgments."]
Add New Comment
Viewing 4 Comments
Thanks. Your comment is awaiting approval by a moderator.
Do you already have an account? Log in and claim this comment.
Do you already have an account? Log in and claim this comment.
Do you already have an account? Log in and claim this comment.
Do you already have an account? Log in and claim this comment.
Do you already have an account? Log in and claim this comment.
Add New Comment