Reinventing Discovery - Michael Nielsen [34]
Converting Individual Insight into Collective Insight
In addition to coordinating attention, the MathWorks score also served the important purpose of helping turn the insights of individual participants into collective insights held by the entire group. Every time someone had an idea that improved a program, this was reflected in their score, making the value of their new idea immediately apparent to all participants. For collaboration to succeed, there must be some way of converting individual insight into collective insight. In other words, the collaboration needs to know what the collaboration knows.
Kasparov versus the World shows what happens when a collaboration only imperfectly converts individual insight into collective insight. As we’ve seen, the World Team relied on Irina Krush and her colleagues to identify and publicize the best ideas of the World Team. Without Krush’s skill at evaluating and comparing analyses, the World Team would likely have done far worse at aggregating the best ideas. Of course, even though Krush and her colleagues put in a mighty effort, their manual approach wasn’t as fast or objective as the automated scoring in the MathWorks competition. As a result, much of the available expertise on the World Team was squandered. Many experienced chess players participated on the World Team, and while some enjoyed the experience, others felt alienated, believing their insights were lost in the general noise of discussion. Years after the game, one participant wrote in an online forum:
If anything in my life that I’ve participated in that I could label as a perfect example of how a community should NOT solve a problem, it was the KvW match. (which I particpated in heavily and am a master (fide [chess rating] 2276)).
Such disaffection occurred because Krush and a few colleagues were manually integrating the best ideas of thousands of people. Their efforts were remarkable, but of course they could only do the job imperfectly. This caused occasional frustration on the World Team, and almost certainly some missed opportunities. This is a general ule: the more effectively a collaboration can convert individual insight into collective insight, the more effective the collaboration will be.
In fact, the World Team’s system for converting individual insight into collective insight broke down badly at a crucial point in the game. As I mentioned earlier, until move 51 the game had seesawed back and forth between Kasparov and the World, with neither side gaining a decisive advantage. By move 51, Kasparov was in a slightly stronger position, and the World Team was fighting for a draw. Unfortunately, at move 51 a member of the World Team by the name of Jose Unodos claimed to be able to break Microsoft’s voting system, and to have stuffed the ballot in favor of a move that he personally liked, but that was not considered a strong move by Krush and most of the other top World Team players. Jose Unodos’s preferred move won the vote, the first time since move 9 that Krush’s recommendation wasn’t played by the World Team. The event helped tip the balance of the game in favor of Kasparov, and damaged the World Team’s morale. Eleven moves later, Kasparov won, in a sad end to one of the great games in the history of chess. When a group’s ways of converting individual insight into collective insight break down, collective intelligence no longer functions. In the next chapter we’ll see that in some fields such breakdowns impose fundamental limits on collective intelligence.
CHAPTER 5
The Limits and the Potential of Collective Intelligence
Collective intelligence is not a problem-solving panacea. In this chapter we’ll identify a fundamental criterion that divides problems where collective intelligence can be applied from problems where it cannot. We’ll then use