Saturday, December 25, 2010

Group IQ

Not only do we have individual IQ, but group IQ as well.

Despite all the well-meaning diversification of measures of human intelligence into such things as emotional intelligence, spiritual intelligence, and the like, (even basketball intelligence!), the big kahuna remains IQ- general cognitive intelligence. This single measure correlates (to various degrees) with a wide variety of real outcomes, such as academic performance, income, social status, longevity, and health. Putting aside the question of how some tests for IQ might be culturally biased and thus merely entrench the social priviledge they are supposedly used to mitigate, IQ has been a scientifically interesting and rigorous measure for a long time.

One would not necessarily expect that the ability to recall words, or do math & logic problems, recognize patterns, or fulfill other odd questions put by intellegence tests would indicate general capacities to succeed in life. But they do, which indicates that our brains, for all their modularity and complex architecture, have a general and variable capability to grapple with problems, in life as well as on tests. Problems are the real test of intelligence, and not just for humans. Outside my window is a bird feeder which I have had to re-engineer several times after local high-IQ squirrels "solved" its puzzle. It has been almost like an episode of Caddy Shack!

A recent paper shows that individual IQ (also called "g", for general intelligence) doesn't account, however, for performance of small groups. The experimenters derive a new measure, called collective intelligence, or "c". As one can intuitively appreciate, groups have a dynamic of their own, depending often more on how sensitive people are to each other and how well they draw out each other's abilities and virtues, rather than how intelligent each individual is. Getting this dynamic right is the holy grail of management books, courses, and gurus- for business especially, but also for government, nonprofits, academia, juries- all types of groups, really.

This study was done on seven hundred people from the general public, put into experimental groups of two to five and faced with a standard battery of diverse small group problems. The researchers then looked at whether groups tended to perform in correlated ways over disparate tasks, and at what factors accounted for that performance. They found that some groups performed better than other groups across the board, on tasks ranging from checkers and visual puzzles to negotiation, indicating that there is a general collective intelligence factor. To ask how many variables were at work across the different groups and tasks, they performed a factor analysis, which pointed one hidden property that accounts by itself for half the variance in the data ... which they call "c".
"Empirically, collective intelligence is the inference one draws when the ability of a group to perform one task is correlated with that group’s ability to perform a wide range of other tasks. This kind of collective intelligence is a property of the group itself, not just the individuals in it."
But what is "c"? Factor analysis points to hidden variables, but doesn't say what they are, or what they are composed of. One thing it is not, surprisingly, is individual intelligence, which correlated weakly with group performance, whether measured as average group member intelligence, or maximum group member intelligence. The figure shows the relative correlations:

The researchers combed through a few other obvious candidates for measurable group and individual characteristics that might correlate with "c", and which might then be compiled into a proxy/predictive measure for that major factor in group intelligence. They mention looking at the following:

  • Group cohesion
  • Group satisfaction
  • Group motivation
  • Individual personality / temperament profiles
  • Individual psychological security profile
  • Average social sensitivity (which can be measured)
  • Proportion of females
  • Variance of speaking turns- did everyone participate?

The first three, surprisingly enough, showed little correlation, while the last three had high correlation. To put it in numbers, individual intelligence correlated with group performance at 0.18, average individual social sensitivity correlated 0.26, proportion of females correlated 0.23, and variance of turn taking correlated -0.41, (high variance means that one person dominated the group). They mention that the correlation with females in the group was probably the same effect as the social sensitivity measure, so they may be measuring the same characteristic.

Turn taking variance is a group characteristic, thus not the kind of characteristic that can be predicted individually, making this one difficult to apply in general way to group performance prediction, even though it represented by far the highest correlation. Additionally, it could be just another (practical) expression of that same social sensitivity that was measured in other proxy ways.

More interesting was the role of female presence and the correlated social sensitivity and group performance. Now who would have predicted that?!

"Padoa-Schioppa’s proposal that the euro would unite the economies and peoples of Europe is turning out to be true. The common unifying element is the entrenched unemployment that the system has delivered which will define the European landscape for years to come unless nations take the only sensible step and exit the defunct and unworkable monetary system."
and ...
"Now we tolerate high levels of unemployment without a clear understanding of the magnitude of costs that that policy position imposes on specific individuals and society in general. The neo-liberal publicity machine has been running full bore for a few decades now to convince us that unemployment is not a major issue that should be addressed."
and ...
"As an aside, it is true that US is losing ground in global terms and I agree with Johnson on that score. But the US budget deficit is not the reason for the loss of traction.
You cannot go around shooting up places illegally; engaging in wars that actually make the world less safe; treat your own people like dirt by forcing them into joblessness for extended periods; tyrannise and torture prisoners of war in illegal detention camps; and more … and then lecture the rest of us about how wonderful and free and dynamic your nation is. The rest of the world (bar the Eurozone) is moving on and is smarter than that!"


  1. Burk - nice post! I'm interested to know whether there have been studies which look at the dynamics of larger groups. As we all know (thanks to Men In Black), "a person is intelligent; people are stupid." In a large group, "mob mentality" often rules, which could be stated in simplistic form as: the effective average intelligence of the group is lower than the actual average intelligence of the group, or, there is a negative correlation between group size and group intelligence (if one doesn't consider it to be a threshold effect).
    Was there a correlation between group size (in the small study groups) and group intelligence "c"?

  2. Hi, Kelley-

    Does size matter? An excellent question! I was just reviewing a paper here, not studying the field in depth. I doubt that larger groups have been studied in any kind of systematic way- there you get into political groups, and quickly beyond the capacity to do any kind of controlled science.

    They don't mention testing size as a factor, which is indeed surprising. The variables they mention were quite few. Are juries an optimal size? All this is pretty interesting with all the social media .. what is the optimal number of facebook "friends"?

    But I'd speculate that there are huge differences in the effectiveness of large vs small groups. Transactions become more one-way than mutual, and social intelligence and sensitivity become more a function of reading the zeitgeist and tapping into hypnotic suggestion than reading someone's eyes and expression, and engaging in retail negotiation.

    This is where religion really comes into its own. How best to align the motivations of large numbers of people? One would want to deploy numinous, meaningful symbology. One would want to offer goals of momentous significance that reach beyond the mere utilitarian.

  3. You're right, it's certainly difficult to study large groups in a systematic and controlled manner. But this ability within a large group of "reading the zeitgeist and tapping into hypnotic suggestion than reading someone's eyes and expression, and engaging in retail negotiation" is a form of "intelligence" in its own right. I suppose we'd more like to refer to it as manipulation... but without a certain "manipulation quotient," the scheming wouldn't be effective.
    I'm surprised there's nothing in the paper about the effect of group size: having five people in a group is dynamically a lot different than only two! I've seen some work on the maximum group size a human is capable of dealing with, but nothing to express the dynamic of such a group in terms of an intelligence quotient. It's fascinating, though.

  4. There have been studies done on ideal group size, which (according to these studies) is 150. What I've read about these studies looks at social network size (not just online but in real life) and also at working environments (like office size or factory size). I read about this in the Malcolm Gladwell book Tipping Point; he provides further references in the book on the studies and other analyses. I wonder if there is real value in utilizing small groups (of say, 2-5 people) within an environment of 150. If 150 is the ideal size for maximizing productivity and efficiency (and general social satisfaction), is there further usefulness for small groups?


  5. Meghan, I'd seen what must be similar (or complementary) studies which indicate that 300 is the maximum group size to retain cohesiveness/effectiveness (the number varies for different species of primates, with humans having the highest value). From experience, we know that these larger groups tend to break down into smaller ones (in the few to ~dozen person range) in order to accomplish tasks, then "reporting back," if you will, to the larger group (think of the different sections in an orchestra or band). If we can naturally arrange ourselves into this efficient model, it could still be asked - has anyone actually quantified its effect, comparing it to other patterns of group behavior?