I bring here a version of an article published on the website of Emotools almost a year ago: “¿Qué factores predicen que un grupo sea más inteligente?”. Perhaps it seems an old article, but it is worth because it fits the main object of this blog and complements other entries. It was one of the issues most cited in Collective Intelligence MIT Conference 2012 held in Cambridge (Boston) two years ago. This research was done by a team formed by Anita Woolley (Carnegie Mellon) and Christopher Chabris (Union College/MIT), among others, whose results were published in the journal Science with a significant media impact.
By explain in a few words, the challenge was to find out if there are any factors that measure and explain the “intelligence of a group” as an ability to solve tasks by a group in the same way that there is an “Intelligence Quotient” (IQ) that estimates the degree of individual intelligence. Hence was born the so-called “C-factor“, which is the counterpart of the IQ coefficient but at a group level.
The experiment was conducted with 699 volunteers, who formed 192 groups of two to five people each. These groups were assigned cognitive tasks of different nature and chosen from the taxonomy of Joe McGrath on group tasks (See article in PDF Susan G. Straus “Testing a Typology of Tasks“). After compiling the results of work packages, these data were compared with those of individual intelligence tests previously conducted with the participants and other attributes of the group members (e.g., gender, age, motivation, sociability, etc.).
The most important conclusion was that neither the average intelligence of the group (average of the individual IQ coefficients) nor maximum intelligence (the highest IQ among individuals in the group) was a good indicator to predict intelligence of a group (C-factor), i.e.: “Just getting a lot of smart people in a group does not necessarily make a smart group“.
On the other hand, some factors that are typically used to explain the intelligence of a group as the “group cohesion”, “group happiness” or the “degree of enthusiasm” had a much smaller impact than previously thought, while the three elements that best predicted group intelligence were: 1) the average degree of social skills of members, 2) a more distributed conversation within the team, 3) a greater number of women. The combined effect of these three ingredients explained over 43% of the observed variance in group’s performance, and that combination gave rise to the so-called “C-factor”. I will discuss in more detail each of these three elements:
1.- Social skills of members:
One of the most compelling factors was the “average degree of social skills” (“average social sensitivity“) of those who formed the group, i.e. empathy, openness and sensitivity that each member had for others. This was measured using the Test “Reading the Mind in the Eyes“, which is to interpret what a person is feeling at each moment from images of the eyes of different moods, starting from the premise that while more able we are to “read the eyes“, more sensitive we are (here’s the test).
2.- Distributed participation:
According to the study, while more distributed the participation was (i.e. less inequality or variance in the degree of involvement among members), the group was smarter. In fact, this variable predicted even better (“participatory variance“: -0.41) than the “social skills” (+0.26). The “degree of participation” was measured not only by the frequency in which they intervened orally, but also for signs of non-verbal communication indicating how each person was engaging in the collaborative process. This was measured by a sensor device called “Sociometer” invented by Alex Petland (I explained the device in this article).
3.- Number of women:
The investigation revealed that the increase of women in the groups significantly improved the C-factor, or in the words of Thomas Malone: “Groups with more women tend to be smarter than groups with more men”. That is: “More females, more intelligence” (NOTE: ceteris paribus). The latter should be qualified because its effect may be more a result of the first factor than an independent variable in itself, because women tend to have more social skills than men, and this feature reinforces the first factor of “social skills” of the group.
In addition to these three variables, the experiment indicated that the collective intelligence (C-factor) improves with increased cognitive diversity of the group but there is a point at which such diversity can begin to adversely affect the IC. The graph that links both factors would be something like an inverted U, but very wide. There seems to be a “point of optimum diversity“ for the intelligence of a group because in this, as in everything, the excesses are also paid.
Moreover, if we compare two groups, one with a higher C-factor than another, we see that as we increase the number of rounds and interactions within them, the group with the highest C-factor learns much more than the lower, although the latter had started knowing more. This observation reinforces the idea that smart teams facilitate the learning process faster and better.
In my opinion the predictive validity of the C-factor as an indicator of intelligence of a group is still debatable and it deserves much more research, but at least it helps to get to know what features of a group influence on the collective performance of a team. An interesting lesson is that in the creation of “smart groups”, social skills such as empathy or “participatory engagement” (wait, listening, sharing, noninterference) may matter more in the effectiveness of group than individual brilliance of its members. In this direction it is already examining the possibility of “rapid tests of social sensitivity” to determine the optimal composition of smart groups, or increasing the “C-factor” of a team through specific training to improve social awareness of its members and the ability to generate models of distributed participation.
Note-1: The image of the post belongs to the album of alles-schlump-f in Flickr
Note-2: Read this post in Spanish (Lee este post en Español)