Really difficult challenges require extreme performance by great teams. In sports we celebrate the performance of extremely effective teams. In music we celebrate extremely accomplished bands and groups. Images of extreme performance teams in our culture include photos of mission control for Apollo 13 (right).
Familiar proverbs suggest basic elements. “Two heads are better than one.” “Many hands make light work.” But what are the patterns of a high performance team? What are the dimensions of extreme performance?
Team Chess: Extreme Performance in Closed Worlds
In widely reported competitions in 1996 and 1997, a specially programmed IBM computer system, Deep Blue, defeated Gary Kasparov (right), the reigning world champion of chess. This was generally seen as evidence that artificial intelligence could be superior to human intelligence. More interesting in the following decade has been Kasparov’s shift to playing chess using a computer as a partner. In traditional chess tournaments, playing as a team or playing with computer assistance was regarded as a kind of cheating. Kasparov created a new kind of freestyle tournament where all that matters is winning the game. Any kind of team or augmentation is fair play. Hundreds of games in such tournaments have now been played and studied.
In his review of Diego Rasskin-Gutman’s book Chess Metaphors published in the New York Review of Books, Kasparov reported on a startling surprise in one of these free style tournaments. In the usual course of events, teams of humans plus computers defeated all challengers including the strongest computers. But in 2008 something different happened:
“The surprise came at the conclusion of the event.
The winner was revealed to be not a grandmaster with a state-of-the-art computer, but a pair of amateur American chess players using three computers at the same time.”
What happened was that a team of weak human players plus a machine using a better process beat all comers including teams with strong human players plus a machine with an inferior process.
The table on the right summarizes the relative advantages in the winning team. As Kasparov says, on these teams he plays chess differently. He does not need to devote much attention to avoiding simple mistakes, and he devotes more time to strategy.
Game Changers: Implications for Open World Problems
Chess is an example of what artificial intelligence researchers call a “closed world” domain. In closed worlds, the rules are established and known. In a closed world like chess, the challenge in winning is in finding the best moves in a huge combinatorial search space.
Real world or “open world” problems are not so simple. New rules of the world — new kinds of moves — can be discovered at any time. The combinatorial challenge of a huge search space is still there, but there is also a search space of discovery and invention. New kinds of moves can always be discovered and winning requires both discovering them and learning. Winning and even survival can depend on rapid adaptation.
Game changers are innovations that redefine the competition. The iPhone™ (right) is such an example in the world of computer electronics. When the iPhone was announced, every other team that had a phone product in the works had to start over. The new phone design so radically changed the game that previous designs became obsolete.
Augmenting Team intelligence
What are the possibilities for human-computer teams in open worlds?
As in the table about human-computer teams on the left, computers have an advantage of speed. In closed worlds, computers are faster at generating and evaluating moves in combinatorial search spaces. In open worlds they are also faster at processing “big data”.
Both advantages leave openings for humans. In open worlds, the rules are not complete so there is room for innovation. No matter how big a data base is, there is always information about the world that is not completely represented. There are always unknowns and potentially surprises. (Rumsfeld at right).
In 1973, Chase and Simon suggested the chunking hypothesis, that the abilities of advanced chess players to copy and recall positions was attributable to the storage of thousands of chunks or patterned clusters of pieces. They estimated that at the highest levels of performance, this corresponded to 10,000 to 50,000 hours of practice. This model of acquisition and use of expertise has been studied now for many domains.
Common sense from 100,000+ hours of living may enable human members to sustain the advantages of augmented teams.
A small group of us is looking at the cognitive activities of teams on open world problems and ways to augment teams with computers to achieve extreme levels of performance. Our project on augmented team intelligence is studying system architectures for augmentation and kinds of activity roles for people and computers on extreme performance teams for open world tasks.
(This post is drawn from a paper by Hoda Eldardiry and me presented at the Collaborative Analysis and Reasoning workshop at the Collaborative Systems and Technologies Conference on May 22, 2014). Thanks to Ed Feigenbaum and Dan Bobrow for comments on earlier versions.