Doesn’t mean that the tool doesn’t work, it just means that you have to be careful using it. Talk about hitting your finger with a hammer!

In this debate, one aspect appears overlooked: communication. If you don’t have the adequate communication policies/ environment, how does one reap the rewards from all this multidisciplinary goodness?

Again, the article stands out in its own but it made us think of the multidisciplinary approach “debate”. We’ve always tried to build our teams with this in mind, looking for complementary skill sets/ experiences/ backgrounds. Nothing new there. But please bear with us as we follow the “links” this article made us follow.

The diversity prediction theorem by Scott Page states that:

“Collective error = average individual error – prediction diversity” (as stated in Michael Mauboussin’s Think Twice, Chapter 3, page 46)

Some types of problems are better suited for experts and some for “crowdsourcing”. Either way, the theorem above applies. For instance, when diversity doesn’t improve the odds of solving a problem, you can focus on decreasing the first part, i.e. decreasing average individual error or, if you prefer to see it this way, increasing “ability”, as Mauboussin calls it.

But if you think proactively and seek counter-evidences and multidisciplinary approaches yourself, it’s as if you’re increasing the “diversity” part of your own expertise/ skill set/ tool set. At the very least, you could become less prone to fall into the traps of biases and etc.

In the book, Mauboussin recommends reading Scott Page’s book (“The Difference”) and his own text on the subject.

A review of “The Difference” at the Crooked Timber blog mentions in the 8th paragraph the problem with one of the assumptions of the diversity prediction theory (just to be clear, Mauboussin deals with it in his book): it assumes that the diverse crowd interacts/ communicates in a way that the diversity “bonanza” can take place. As we at IP know, that is often far from reality.

Another approach to the same problem was discussed by Julia Kristeva: “One cannot be an amateur, or decide one day ‘Let’s be interdisciplinary’. A university may decide to develop in that direction, but what matters is that each researcher finds and establishes some complicities with other researchers so that interdisciplinarity comes from the base of the pyramid and works its way up. One can only benefit from interdisciplinary practices if researchers meet other researchers whilst learning how to discuss both their competencies and the outcome of their interaction; therefore contributing to the exposure of the risks inherent in an interdisciplinary practice… the first obstacle is often linked to individual competencies coupled with a tendency to jealously protect one’s own domain. Specialists are often too protective of their own prerogative, do not actually work with other colleagues, and therefore do not teach their students to construct a diagonal axis in their methodology.”

At IP, we’ve always taken steps to address this and this blog is part of these efforts. For instance, our team’s interaction in our Intranet today is the source for most of our posts here – this specific one was “built” from the contributions of 4 different members.

The way MIT built its incentive system to recruit people to its “team” is part of the story, of course. They were clever enough to mix both “ego-boost” and “cold, hard cash” incentives (since they would donate some of the proceeds to charity). Other teams apparently had either just the feel-good or just the greedy incentives. Of course, the losing teams complain that they had been working for a while when the MIT Team came out of nowhere and, given the famous backers, got a lot of traditional press attention, giving them a leg up on recruiting “balloon spotters”. Well, it seems good media contacts do play a major role in this part of the challenge, so DARPA probably got a lesson from that as well.

Speculations abound on what DARPA could have learned from this, but whether this will help the US capture the “terrorist that can’t be named” is beyond our scope. What we’re interested in is the wealth of data on crowd-sourcing and information dissemination generated by this contest. We’ve posted before on the subject of Crowdsourcing and the implications on Media, Online Services, Search, Hyper-local advertising, Retailing, Services in general (links below). Here’s hoping the MIT Media Lab team will write a comprehensive paper on this challenge and shed some light on this trend.

LINKS:

With lure of cash, MIT wins Pentagon prize – NY Times (free registration required) on the contest winners – Dec. 6th ’09

“Crowdsourcing” revisited – Our own take on crowdsourcing and its potential shortcomings – Nov. 19th ’09

Netflix’s smart crowdsourcing initiative – Our take on the Netflix contest – Sept. 22nd ’09

MIT’s 2-day course on crowdsourcing – Classes in 2010 for this $2,600, 2-day course called Intelligent Organizations: Collaboration and the Future of Work.

The point here is, on the company side, to remember Warren Buffett’s “institutional imperative” analogy. A manager’s job is not to keep up with buzzwords, after all. Crowdsourcing is a new thing, therefore relatively untested, and it certainly has its optimal applications and its limits – which probably haven’t been found yet. It’s just another tool available and should be subject to the same rigorous analysis before the company starts using it.

Links:

The crowd is wise – when it’s focused – NY Times, July 2009 – here’s the link to a MIT paper they quote called “Harnessing Crowds: Mapping the Genome of Collective Intelligence”

Wired’s 2006 seminal article on Crowdsourcing – by Jeff Howe, who later wrote the inevitable book version.

Jeff Howe’s blog on the subject – Pretty much dead, but it has links to other Crowdsourcing stories.

But that’s not all that matters: is crowdsourcing then a substitute for companies’ R&D demands? Does it always work? How open do you really want to be? Some people are looking at it and saying “hold your horses”. That’s always a good advice in our book.

As for Netflix, they’re happy enough to have a Netflix Prize 2 on the works. Very smart: it keeps the ball rolling and takes advantage of the community they built. The press coverage doesn’t hurt, either.

But there’s more. We’d argue that the benefits of the Netflix Prize in terms of research in statistics/ data mining/ etc. in large data sets may be felt for years. It’s interesting to consider this in terms of efficiency gains for many different types of companies. It’s something companies all around the world should be paying attention to.

Here’s a video interview with Reed Hastings, Netflix’s CEO, at the time of the award.

Links:

Crowdsourcing page on Wikipedia

Jeff Howe’s classic article on Wired, his book and his blog

The teaser for Netflix Prize 2

Winning the Netflix Prize – NY Times, October 2008 (huge article with a video that apparently can’t be embedded here).