ARL funding develops concepts in collective intelligence

February 14, 2011

ARL funding develops concepts in collective intelligence

By Dr. Joseph Myers, U.S. Army Research Office Information Sciences Directorate

The newly-developed concept of collective intelligence could someday predict how an existing group or organization will perform on new tasks, help select group members from a population to form maximally-functional teams, and eventually do the same for groups made of humans and machines working together.

General intelligence measurement in individuals is essentially factor analysis, seeking to explain the variability in performance in a number of diverse cognitive tasks in terms of a smaller number of unobserved variables, the most significant of which is labeled general intelligence ("g").

ARL-funded principal investigators (PIs) at the Massachusetts Institute of Technology and their colleagues at Carnegie Mellon University are developing analogous measures for groups. Through performance measurements of many diverse groups of two to five people on a diverse set of cognitive tasks, the researchers have identified a single significant factor that explains most of the variation in group performance; they have labeled this factor collective intelligence ("c").

These researchers have found that group characteristics most significantly increasing collective intelligence are: social perceptiveness, evenness of interactions among members, and higher proportion of female.

Experiments, and thus results, are so far limited to small face-to-face ad hoc human groups; researchers plan to soon investigate extensions to larger, functional human groups. While general intelligence in individuals is difficult to change, this project has shown that collective intelligence can be modified through changes in personnel, motivation, procedural rules, and collaboration tools (e.g., email, wikis, etc).

Potential uses of this new understanding include the ability to predict performance of an existing group or organization on new and different tasks, to predict performance of a not-yet assembled group on a variety of tasks, to select group members from a population in order to form maximally-functional teams, and eventually to do all of the above for human-machine groups as well.

The United States Army Fires Center of Excellence at Fort Sill, Okla. is discussing with the PI the possibility of doing trials involving their Air Defense Artillery teams, and discussions with the Office of the Director of National Intelligence on possible trials are ongoing.

This research has received considerable media attention over the past three months. It has appeared more than 40 times in publications, such as Science, Wall Street Journal, New York Times, NPR, the web version of Scientific American, and other print, radio and web sources.


Last Update / Reviewed: February 14, 2011