Mind & Brain / Memory, Emotions, & Decisions

My proposed rules are a little different from Surowiecki's in that they are framed more from the outside looking in at the crowd. For example, I would argue that a crowd shouldn't be allowed to frame its own questions and that the answers should never be more complicated than a single number or a single multiple-choice answer. I also propose that techniques usually associated with signal processing should be applied to crowds, like gradually changing how fast a crowd can act based on how it is performing.

Maybe if you combined our approaches you'd get the magic seven rules; maybe those could be compressed into a smaller number. Then again, maybe one or both of us are on the wrong track. The problem is that there's been inadequate testing of such ideas. Numerous projects have looked at how to improve specific markets and other crowd-wisdom systems, but too few projects have framed the question in more general terms or tested general hypotheses about how the systems work. What a rich area to study!

There is another potential pitfall of crowd wisdom: the ability of information technology to lock in cultural or behavioral patterns. Suppose you don't set the rules effectively in advance and the crowd acts in an ugly way. If the connections within the crowd are mediated by digital technology, then an engineering challenge stands in the way of fixing the problem.

Internet-based designs like Wikipedia and Digg are enjoying a period of open possibility right now, but that won't last forever. The Internet at present is analogous to that digital Eden of the early 1980s, when personal computers still seemed mercurial and infinitely mutable. Layers of digital design can become locked in place because other layers come to depend on them. The PC, for example, has become a standardized thing with windows, a mouse, a hard disk divided into files, and so on. Our computers could have come out quite differently; the Macintosh was originally conceived without files as we know them today, although it acquired them before it shipped. Today, ideas like files and windows have become so entrenched that they might as well be elementary particles. While those ideas are probably not terribly influential on human behavior, locked-in Internet-based designs could be decisively important.

That's why there is such passion from all sides about the battles over things like digital rights management, digital privacy, and Net neutrality: When life is digital, any battle might turn out to be the end of the war. If a more restrictive outcome gets locked in, it becomes profoundly difficult to reverse. We are spectacularly lucky that the people whose early experiments turned into the Internet conceived of an optimistic open design that happened to get locked in.

The legacy effect might eventually ossify aspects of the Net that are now fluid. There is an astonishing and widespread denial about this process in some corners of the software world, particularly in the trendier domains of open source. The situation might improve in the future, but that will require fundamental improvements in the way we do computer science. I hope it happens. In my professional life, it's the problem that I work on the hardest. I'll talk more about that in a future column.

There is a third empirical problem to tackle, and it is the least comfortable. To what degree is mob behavior an inborn element of human nature? There are competing clichés about human identity: that we naturally and inevitably form into competing packs or that we would refrain from doing so if only we had decent gang-free peer groups in our teens. These theories can actually be tested. The genetic aspects of behavior that have received the most attention (under rubrics like sociobiology or evolutionary psychology) have tended to focus on things like gender differences and mating strategies, but my guess is that clan orientation will turn out to be the most important area of study.

I hope that improved understanding of the problems I've mentioned will come about before the Net can contribute to any large-scale outbreak of bad human behavior. A better and more general model of when the wisdom of crowds functions and when it breaks down will help us avoid Web-based designs that elicit cruel or stupid mob behavior. A better technical approach to avoiding the lock-in effect will help us correct mistakes along those lines if they occur. And a clearer picture of the nastier side of our genetic legacy will help us design information systems to avoid triggering evil behavior.

We need such breakthroughs soon. Based on the ever-growing influence of the Net, my guess is that we have about 10 years to seek out the answers.

Previous Jaron's World columns: