Here is an interesting comment I received about my post on “Wisdom of the Crowd” from Michael.
“This column by Dov and the factors it explores do not translate in any intellectually rigorous way to assertions that a random group of people on Facebook or responding to a post on a website will be wiser than experts. (Wikipedia is a whole different issue.) Random Post No. 7 may be brilliant, but you miss it because Random Post Nos. 8-10 put it down. Or you get misleading information until Random Post. No. 17 nails it. The give-and-take and real-time interaction of a real crowd has merit that can certainly be replicated online but cannot in any way be taken for granted. See: Wisdom of Crowds Talking About Sports.”
Great comment Michael, and here is my response.
The phenomenon of collective intelligence is extraordinarily complex. Let’s consider some of the levels of complexity. At its most basic level is the individual. The psychological aspects involved are the products of a myriad of genes, all interacting with each other, and those intertwined genetic determinants are in turn subject to environmental influences impacting each gene individually, and its interactions with other genes. This complexity boggles the mind. But we can’t stop here. Each individual interacts with other members of a group. The group dynamics adds another level of complexity. Furthermore, we cannot extrapolate results from a small group to a massive one. The phenomenon of “mob psychology” is unfortunately quite common, albeit poorly understood. Add to that modern complications, such as virtual social networks, virtual “friends” on Facebook, and you are tempted to throw your hands up –just too complicated.
How do we tackle such a “hopeless” problem?
Science has faced such tasks in the past, and quite successfully. The most recent example is the study of the brain, which encompasses the fields of neuroanatomy, neurophysiology, neurobiology, neuropsychology and psychiatry. The field seemed so overwhelmingly complex that only philosophers and theologians dared to offer their theories of mind. Needless to say, philosophical musings were just that –musings. Models of the brain were proposed based on computer architecture, the organization of the immune response, quantum mechanics, and some ethereal “soul” or ”vital force”. They lacked the intellectual rigor founded on a body of data, and hence failed to move the field forward.
Painstaking studies on the anatomy and physiology of a single neuron were followed by studies on the neuronal pathways. The simple nervous system of the worm C. elegans served as a model for studies of more complex organisms like the mouse. And using the newly invented neuroimaging techniques finally allowed us to study the higher level of complexity underlying the functioning of the whole brain. We are still at the beginning of unraveling the enigma called the brain/mind, but it took a century to get to where we are now.
The reason I went into such detail is to illustrate how the scientific process works. There are no shortcuts –it is a laborious process of starting with the basics and amassing data.
This is exactly the process that science takes in elucidating group intelligence, with its daunting complexity.
The paper I cited in my last posting (Science on-line, Sept. 30, 2010) did not attempt to find the “big truth” in one fell swoop. Woolley’s team divided 600 test subjects into groups of two to five people, then had each group complete a variety of problem-solving tasks. Afterward the researchers interviewed the groups and each participant. They measured group cohesion and motivation, individual intelligence and personality, and other factors previously associated with group performance. Their analysis found several characteristics linked to group performance — and none involved individual intelligence. What mattered instead was the social sensitivity of individual members, the proportion of women (who tend to be more sensitive) in each group, and a balanced participation of conversation. The methodology and detailed collection of data is well thought out and careful; the conclusions are intellectually rigorous, with one possible quibble: Wooley equates social sensitivity with the female gender. Many studies have shown that indeed females are socially more sensitive than males, but still –it’s a bit of a logical leap to a priori assume that it is valid in this study.
As to the argument that Facebook-like groups may not support the thesis of group intelligence, we have no data one way or another. The argument sounds plausible, but needs to be studied. In any event, the paper at hand dealt with small groups, face-to-face interactions, not on-line communities.
On-line Group Wisdom
What do we know about the intelligence of on-line communities? No empirical data that I know of are available. But several intriguing observations give us some clues.
One example is on-line betting about election results. The accurate projections were not given by polling organizations (remember the fiasco of the 1998 election projections based on exit polls?)-an internet betting site was by far better. Most curiously –the British betting site was even more accurate about our election results than the American one. What’s the secret? I think the answer is that participants couldn’t just vote their prejudices; they had to put money on the line. Apparently this tends to concentrate the mind; it also filters out partisans trying to sway the results of the poll. In any event, the large number of participants would have rendered any such attempt pointless.
A more familiar example: the stock market. Here millions of people are daily rendering judgments on economic issues –and are putting billions of dollars where their collective mouths are. Are these people all economists? Not by a long shot. What they are doing is betting on what they think “the other guys” are betting on. And this requires a degree of social intelligence. Is the group intelligent? The market sometimes seems to act irrationally, tending to extremes of optimism and pessimism. But invariably, it eventually “reverts to the mean”. Is “The Market” wiser than the experts? Absolutely. Burton Malkiel (Princeton), Martin Feldstein (Harvard), Nouriel Roubini (N.Y.U) and assorted other luminaries developed intellectually rigorous models of the economy –and then failed to predict any of the bubbles that afflicted our economy. Daniel Kahneman and Amos Tversky, two psychologists, took an empirical approach and talked to people who actually made decisions about their own money, based on their intuition, not economic models. This empirical approach amassed a volume of data and established the field of Behavioral Economics (for which Kahneman received the 2002 Nobel Prize in Economics; Tversky died a few years earlier). You can see this “intuition” in action every day, in the successful traders who developed a sixth sense about the market “mood”. Few of them are intellectual giants –but they all possess extraordinary social sensitivity.
Having said that, I’d be the first to admit that on-line communities have not been rigorously studied and we know little about them. It will take hundreds of studies like Wooley’s to finally unravel the puzzle of group intelligence.