Edited transcript of remarks at the official launch of the
MIT Center for Collective Intelligence
October 13, 2006
It now falls to me, at this point in the program, to give you an overview of what collective intelligence is, in the first place, and what we’d like to do about it. The working definition of collective intelligence that we’re using is that collective intelligence is groups of individuals doing things collectively that seem intelligent [Q.: mais comment se reconnaissent-ils?].
Now, if you think about it that way, collective intelligence has existed for a very long time. Families, companies, and countries are all groups of individual people doing things that at least sometimes seem intelligent. Beehives and ant colonies are examples of groups of insects doing things like finding food sources that seem intelligent. And we could even view a single human brain as a collection of individual neurons or parts of the brain that collectively act intelligently.
But in the last few years we’ve seen some very interesting examples of new kinds of collective intelligence:
- Google, for instance, takes the collective knowledge created by millions of people making websites for other purposes and harnesses that collective knowledge--using some very clever algorithms and sophisticated technology--to produce amazingly intelligent answers to the questions we type in.
- Wikipidia, at another extreme, uses much less sophisticated technology but some very clever organizational principles and motivational techniques to get thousands of people all over the world to volunteer their time to create an amazing on-line collection of knowledge.
- In just a few minutes, you’ll hear from Alph Bigham the CEO of a company called Innocentive [est-ce vraiement innocent ?] which lets companies with difficult research problems harness the collective intelligence of thousands of scientists in a network all over the world to help solve those problems.
- A lot of companies today--Hewlett Packard, Eli Lilly, Google and others--are now beginning to use things called prediction markets [mais qui paie quoi à qui et comment ?] where people buy and sell predictions about future events (like sales of their products) in ways that leads to more accurate predictions in many cases than traditional market research or polling or other techniques.
Now, I think these examples are just the beginning. With new information technologies—especially the Internet—it is now possible to harness the intelligence of huge numbers of people, connected in very different ways and on a much larger scale than has ever been possible before. In order to take advantage of these possibilities, however, we need to understand what the possibilities are in a much deeper way than we do so far.
So, I think the time has come make collective intelligence a topic of serious academic study. And that is our goal in the Center for Collective Intelligence.
The key question we’re using to organize our work is: How can people and computers be connected so that collectively they act more intelligently than any individual, group, or computer has ever done before?
In order to answer that question, I think at least three types of research are needed. The first is just collecting examples or case studies [Malone : mais tu ne change jamais de méthode ?]. I think there are going to be a lot of natural experiments going on in the next few years, people trying lots of interesting things--with or without us. But I think that we can help the world learn from its experience with all these natural experiments by systematically describing and collecting examples of interesting cases of collective intelligence.
For instance, Eric von Hippel, in the Sloan School, has done some very interesting case studies of how the collective body of users of a product is often a better source of innovation for a company’s products than the company’s own researchers [Ndr : y-avait-t-il vraiement besoin de faire une étude ???]. This kind of case study research is common in business schools, but it is certainly not the only kind of research we need to do.
Another kind of research we need to do is something that is more typical in an engineering school. That is to create new examples of the phenomena we want to study. If you’re an aeronautical engineer, for instance, you wouldn’t just study birds and flying insects, you’d also want to create some flying machines and study how they work. In our case, that means we want to create some new examples of collective intelligence and study how they work.
For instance, Mark Klein in the Center for Collective Intelligence is leading a group of people in a nascent project that hopes to harness the collective intelligence of thousands of people around the world to help deal with the problems of global climate change [& l'Opus Dei ? Qu'en fait-on ?]. We have some specific technical ideas about how to combine computer simulation techniques with online ways of representing issues and positions and arguments that we think may be helpful in this process.
In the process of creating new examples, we hope to advance the state of the art and to learn new design principles not just for the technologies, by the way [BTW], but also for the human, the organizational, the social, and the motivational systems that are needed for these systems to work effectively.
But case studies and creating new examples are not the only things we need to do. I think we also need to do systematic studies and experiments. For instance, in some cases, we’ll find examples of things that work well but we won’t know why from just a case study. So we need to do systematic experiments to help figure those things out. This is the kind of research that would be more often done in a school of science or a school of social science. For example, Sandy Pentland (in the Media Lab), Drazen Prelec (in the Sloan School), and Josh Tennenbaum (in the Brain and Cognitive Sciences department) are all doing different laboratory experiments about different ways of helping groups make predictions more effectively [c'est joli la diversité !].
But these three things--case studies, new examples, and systematic experiments--are not the only things we need to do. We also need new theories to help tie all these things together [ça devient compliqué ...]. I think that is especially important in the case of collective intelligence because there’s now a lot of hype and prejudice going around about collective intelligence [de qui parle-t-il ?].
On the one hand, there are people who think that collective intelligence is magic, and if you just add it, it’ll make everything wonderful. For instance, there is a book called The Wisdom of Crowds by James Surowiecki who--by the way--does not believe what I just said. But many people who’ve heard about his book do believe it. They think that just doing things “collectively” will make everything great.
On the other hand, there are people who are prejudiced against the very notion of collectiveness and decentralization. Very recently for instance, there have been a number of people who’ve looked at the success of Wikipedia and pointed out ways in which is not perfect. And then, based on that, they have argued that nothing without central control can ever be successful.
Now, I think both of these extremes are equally wrong. Sometimes collective intelligence is good; sometimes it isn’t. Sometimes it works, and sometimes it doesn’t. A very important part of our goal is to help put a more solid scientific foundation under the claims in this area.
Fortunately, we don’t have to start from scratch [scratch] in doing that. There’s already a lot of good work that has been done in many fields, including psychology, organization theory, artificial intelligence, brain science and others. Part of what we want to do is to help organize the work that has already been done.
But even if we had already organized all of the results of all of the previous research, there would still a lot to learn. New technologies are now making it possible to organize groups in very new ways, in ways that have never been possible before in the history of humanity. And no one yet understands how to take advantage of these possibilities.
We certainly don’t have all the answers yet; we’re just beginning to ask the questions [ça c'est prudentiellement bien prudent, bravo!]. We hope we can make a contribution just by helping to frame the questions better [je suis d'accord : le plus important c'est de poser les bonnes questions ... c'est déjà fini la lecture ...]. We’re pretty [pretty] sure we can have a lot of fun trying. And we hope that in the long run the work we do in this center will help contribute to scientific understanding in many different disciplines and help us understand new and better ways to organize businesses, to conduct science, to run governments, and--perhaps most importantly--to help solve the problems we face as society and as a planet.
