One thousand brains are better than two

Common wisdom teaches us that individual ingenuity is essential for advancing human knowledge. However, it is becoming increasingly evident that the opposite can be the case: a large group of properly organized people can accomplish intellectual tasks impossible for individuals or small groups. I first realized the power of this idea when the New York Times Magazine published its Year in Ideas: 2009. The article “Massively Collaborative Mathematics,” describes an experiment performed by Cambridge mathematician Timothy Gowers. Professor Gowers challenged readers of his blog to attempt to make progress on the Density Hales-Jewett Theorem, a previously unsolved mathematical theorem. Six weeks later, the network of collaborating mathematicians had solved the problem. Using networks of people to solve math problems has not stopped there, either. At the website MathOverflow.net, mathematicians can ask each other research-level math problems -- problems that they are working on which they would like to ask others’ opinions about.

Mathematicians are not the only people who are harnessing the power of social networks to solve academic problems. While computers are still having a tough time telling the difference between a picture of a dog and a picture of a cat, humans seem to have an innate ability to recognize and identify visual patterns. A recent article in Nature News discusses several projects which are harnessing this ability: Foldit invites users to strategize about folding proteins, Stardust@home lets people examine pictures for interstellar particles, and Galaxy Zoo centers around classifying features of galaxies. It’s worth a read, as it gives a nice history of distributed science projects and gives an interesting analysis of the emerging field of citizen science

To be fair, it’s not as though throwing people at a problem is always sufficient to solve it. The experiments highlighted above share a common feature: the ability to make incremental progress. For example, solving math theorems requires the analysis of many possible solutions, most of which are wrong. By cooperating, you can ensure that methods which don’t work aren’t tried over and over again, streamlining the solution process. Similarly, in dataset classifications like Galaxy Zoo, the important thing is going through huge numbers of datapoints; each classification brings you one step further.

The idea that collaboration can take you to places you wouldn’t (or couldn’t) otherwise go isn’t new. But as science is currently performed, many experiments are done either completely alone or with a very small number of collaborators. As we learn more and more, collaboration may become even more essential. The amount that any one person can know is limited, but by discussing ideas and thinking with a group, we can overcome individual gaps in knowledge and make further and faster progress than would otherwise be possible.

I anticipate that the ability to harness networks of people to solve problems will be an increasingly useful skill as the technological infrastructure to support such large-scale collaborations increases. If used properly, large-scale collaborations may bring us to many new and exciting places. The trick to successful distributed science will be two-fold. A scientist must be able to first identify whether his or her problem of interest can attain an incremental solution, and must secondly be able to convince others that it is worth working on. This could be in the form of designing a game, like the creators of Foldit, or by enlisting other interested scientists through on- or off-line networking. It is certainly true that not all problems can be solved this way, but I look forward to seeing what new breakthroughs are made through distributed research.