Jump to: navigation, search

WE measure Collective Intelligence as the difference in quality of a group solution over the best independent individual solution in the group.

WE will move toward a semantic action wiki, using its collaborative knowledge base to drive the structured communication and knowledge building process of the collaborative world and decentralized union of all ontologies on the net to realize our collective intelligence in action.



Most everything worthwhile has been accomplished by individuals. Groups are notoriously bad at making good decisions. Groups most often act with CollectiveStupidity or GroupThink. Pure democracy is simply the TyrannyOfTheMajority. Some things may work well with collective intelligence, some things less well. In P.J.Plauger's book "ProgrammingOnPurpose" he describes setting problems to his computing class, splitting the class up into teams of different sizes. For design problems a team of four was productive, the team of twenty achieved nothing. For 'BrainStorming' problems (come up with the most original/off-the-walls suggestions) a team of ten did better than any of the other teams. Plauger wasn't officially teaching the class computing - but the important lesson for them was group dynamics.

How well CollectiveIntelligence works depends very much on how the group structures itself. Our goal must be to discover and communicate the CollectiveIntelligenceStructures that work, and refine these. By doing that we empower the whole group and make it more worthwhile being a part of the group.

See also MeatBall:CollectiveIntelligence.

Some findings of the EIES Legacy era

  • without structure, the larger the group, the worse the decision
  • up to five experts can improve a group decision
  • on-line groups do at least as well as face to face groups
  • face to face groups tend to be dominated by one or two (usually male) participants
  • participants on on-line groups participate equally (gender independent)
  • participants will collaborate only if their privacy and control is protected
  • competitors will collaborate only with respect to established shared principles
  • the ideal group size is two. larger groups can be made most effective by maximizing the effectiveness of individual pairings and decomposing activities into smaller groups down to one on one activities.
  • 25% of on-line group members get addicted to the system and apply it beyond its intended use, another 25% use it effectively, 25% use it due to group pressure, and 25% never use it. Success often demands including the absent.
  • Anonymity and pseudonyms aid forth right communications, particularly in hierarchical organizations.
  • Tailoring a system for a group imparts a feeling of ownership and improves utilization beyond the added value of the tools provided.
  • On-line asynchronous collaborative learning is as effective as the traditional classroom without the time and place constraints.

New findings

  • Among woman, the pleasure centers of the brain are stimulated to a much higher degree by a successful collaboration sharing a greater reward than by winning more personally. (NY Times Science Times 7/23/02)
  • In collaborative simulation games, about 25% will initiate collaboration, 25% will return collaboration in response to collaboration, about 25% require social pressure to collaborate and about 25% will collaborate only under threat. (NY Times Science Times 7/23/02)
  • Bird brains may not be too bright, but pairs of crows exhibit tool making capabilities. (Science times ?/?/02)
  • Contrary to Orwell, Democracy Rules on the Big Animal Farm. (NY Times Science Times 1/14/03)


Other work

I would think that a pod of whales, communicating at high bandwidth, function as a group increadably well. They even work very well with humans with limited communication. We can't speak their language, they have to learn ours. I would speculate this is why a pod beaches themselves after losing a key member. They cannot function alone. Neither can I. -- JimScarver

Defining "Collective Intelligence"

Intelligence is variously defined as "the capacity to acquire and apply knowledge," "the ability to effectively adapt," or simply "the ability to solve problems." But for our purposes it may be useful to list characteristics, capacities or functions that are variously ascribed to intelligence -- problem solving, learning, adaptation, reasoning, prediction, reflection, imagination, etc. -- and then welcome into our domain anyone who is exploring the collective expression of any of these.

Murray Turoff's quantitative definition of Collective Intelligence as problem solving ability, or was it Roxanne Hiltz? I am sure it is in their book, "The Network Nation". In any case we did a whole lot of research at NJIT since 1976 trying to measure it, comparing face to face and on-line systems.

The Turoff-Hiltz definition is the ability of a group to develop better problem solutions than the best individual solution in the group.

This distinguishes collective intelligence from groupthink, which is more the lowest common denominator, mob rule, which tends to degenerate in intelligence as group size increases.

We found people felt less bullied on on-line systems, and were less likely to accept the dominant members viewpoint. (Some on-line subjects had to be physically restrained from physically confronting the objectors to their dictates to the group).

Up to 5 experts can improve a group decision. But even then, consensus is very difficult to achieve on-line. Groups with an elected leader did better. I programmed the "computer leader", which combined the rank orderings of the individuals into a group ordering, entering scripted leadership comments. Groups that elected a leader or had the computer leader reached consensus. Those that had both did not reach consensus, half followed the elected leader, and half the computer leader.

I also programmed a compromise tool, that suggested the minimal compromises which would maximally improve the consensus of the group ordering. Roxanne refused to use it in real groups, she called it Nazi. In my informal tests, however, using this tool, the group reached perfect consensus every time. Students would contribute little of substance and everyone would agree to almost the initial group rankings without any collective intelligence, performing worse than the best individual.

But if collective intelligence is "problem solving" there must be a group decision. The bar should be raised demanding that not only collective intelligence be better than the best, but also better than a simple vote of their individual views. The group solution to a problem should always be at least as good as the best individual solution, or it is collective stupidity. Further, it should capitalize on collective intelligence to improve the best individual solution. Identifying the best individual solution is sufficiently worthwhile as most great accomplishments, a.k.a. problem solutions, are by individuals. Often the individuals are unrecognized in their lifetime, or ever.

The Space program is a great counter example to groups being unintelligent. Science and engineering manage to combine the intelligence of many to accomplish great things. But this is only possible when there is objective criteria for making decisions.

It seems that collective intelligence may require common principals which enable a decision, ideally objective principals.


CollectiveIntelligence SkypeCast

Our first conference on July 23rd was very interesting. We had a few new faces from around the world that stopped in off of the Skype listings.

The next SkypeCast will be on July 29th, 9:00 AM EDT at the link below, running for 3 hours. is where you go to join the conference once we start it.

Download Skype for free at Hope to see you there!

[Note: WE gather that the event referred to above occurred sometime in July of 2006.

SkypeCasts were discontinued in September of 2008.]

Related Links

See DumbAnimals, Consciousness, HumanBrain, NetworkTheory, EvolutionaryGameTheory, SmallWorlds, SmartMobs
See also: GroupWare, SocialContract, AdministratiumAtom, Consciousness, WhyWeDontCollaborate, CollaborativeHypermedia,
Final Project Report for "Coordination in Distributed Group Support Systems",
Distributed Group Support Systems: Theory Development and Experimentation, ExtremeCollaboration

Personal tools