SIEMENS

What’s the promise of collective intelligence (CI)?

Malone: The idea is that there are things that are very easy for humans to do and very hard for computers to do, and vice versa. So the question that CI asks is: “How can people and computers work together to take advantage of what each does best?” That’s the promise. Research shows that even simple computer algorithms can often do a better job of predicting many things than human experts can — things like estimating sales, economic trends, and election results. On the other hand, humans are much better at identifying certain qualitative factors that can influence predictions.

Have you performed experiments along these lines?

Malone: Yes. We are investigating “prediction markets” where participants can buy and sell predictions about future events such as product sales. In our experiments, we let both humans and software agents predict the next plays in an American football game. We found that the agents were significantly more accurate than the humans. But we also found that humans and agents together were more accurate than either alone. The next step will be to build prediction economies. These will include one or more prediction markets — markets for information relevant to an event, and markets for human and machine-based services that can help participants make more accurate predictions.

There are other CI application areas that offer higher probabilities of success. For instance, your web site refers to the possibility of determining whether a growth on someone’s skin is cancerous or not…

Malone: That is a project we would like to do. The idea is that the knowledge needed to resolve the question does not necessarily have to reside in the head of the person standing next to the patient. My guess is that if you had excellent images of the growth that could be transmitted anywhere, and if you had non-physicians who classified such images all day long, then I believe these non-physicians could be — collectively — more accurate than a dermatologist who sees only a dozen potentially cancerous growths per week. Eventually, this very specialized task may be accomplished by an algorithm. But on the way to that goal, we may see humans and machines working on the problem simultaneously.

Are any companies putting this kind of networking into practice?

Malone: Yes. For instance, Amazon Mechanical Turk — a crowd-sourcing marketplace on the Internet — is designed to help software developers build human intelligence into their applications. Programmers can farm out specialized tasks to people, and they can do so in the middle of programs. If, for example, you are writing a program to create a travel directory, you can include a sub-routine that asks people — for a few pennies per task — to read web sites and find hotel phone numbers.

What implications does this problem-solving model have for business?

Malone: One possibility is that much of the work that now gets done inside big companies will be done instead by temporary networks of people and computers. Compensation will range from large sums for solving complex problems to micropayments for things like helping a camera at a loading dock interpret an image when something unusual occurs. There might be on-line lists of situations where human attention is needed, and people could look for the highest paid tasks they are capable of doing.

Can organizations improve their IQ by applying CI to their operations?

Malone: We have recently done some work designed to measure organizational IQ. We gave several small groups a number of tasks and looked at the factor analysis of how they performed. What we found was that, just as is the case with individuals, there is a single statistical factor that predicts the group’s performance on a wide range of tasks. It is conceivable that you could do this at the level of an entire organization. It would be fascinating, for instance, to discover what Siemens’ IQ is and to investigate how we could boost it. We believe that it is eminently possible to change group intelligence. That could have tremendous implications for companies, universities, and governments.