Open source problem solving in science

Linus' Law according to Eric S. Raymond: "given enough eyeballs, all bugs are shallow." In other words, if a large enough community of users and developers has access to (and is using) your source code, even subtle problems will be identified and resolved quickly.

The use of the Internet to create a "massively parallel human problem-solving system" is a powerful concept, as evidenced by such phenomena as the blogger as a source of news, wikipedia as a source of information, and advertising campaigns that solicit user-generated spots. (For more examples, see Jeff Howe's writings on crowdsourcing.)

Now Karim Lakhani of Harvard Business School is looking into whether such techniques can be applied to scientific problems. From a recent article (and interview):

In a perfect world, scientists share problems and work together on solutions for the good of society. In the real world, however, that's usually not the case. The main obstacles: competition for publication and intellectual property protection.

...

What [Karim Lakhani] and his coauthors discovered: "broadcasting" or introducing problems to outsiders yields effective solutions. Indeed, it was outsiders—those with expertise at the periphery of a problem's field—who were most likely to find answers and do so quickly.

He cites a few intriguing examples, including both more traditional "competitions" (with prizes) that have yielded novel solutions, and also more novel approaches such as MathWorks' collaborative programming contest. I am also reminded of NASA's involvement of the general public in analyzing image data from its comet return mission (StarDust@home), and of the nice work at CMU on enlisting people (under the guise of a game) to tag photos. I suspect that there is much much more to be done here.

Comments

[Elizeu Santos-Neto]

This is really a very interesting topic.

I've read a book a year ago about how powerful crowds can be in terms of problem solving. The author try to enlist some properties that a crowd should have to be good for certain class of problems (eg. value estimation).

"The Wisdom of Crowds" by James Surowiecki

[Rosie]

These comments have been invaluable to me as is this whole site. I thank you for your comment.