This brief essay is the tenth in a series addressing the emergence of meaning, by James Leonard.
(Please note: the following material is Â© copyright James Leonard 2006 and may not be used in any way without permission from author)
While in graduate school at the University of Michigan, I had the opportunity to study under John Holland. For years, as an armchair scientist, I had been a fan of his books including Hidden Order: How Adaptation Builds Complexity, Induction: Processes of Inference, Learning, and Discovery, and Emergence: From Chaos to Order. The opportunity to converse with him directly proved even more inspiring than reading his work. One of the many concepts he explains wonderfully (both in person and in writing) is the dynamic between what he calls “exploration versus exploitation.” From chess playing to the stumbling march of genetic evolution, the dynamic between these properties of exploration and exploitation plays out in survival strategies of all types in all systems that have the capacity learn or evolve.
In simple terms, exploitation means going with options already proven to work. On a chessboard, always open by moving your pawn to E-4, allowing your queen and bishop to enter into play early. As a small animal, if a creature is bigger than you, avoid it by running away in a zigzag pattern. Bacteria with a genetic propensity to metabolize glucose sugar will continue to do so for generations if in the presence of a source of glucose. Exploitation can be thought of as conservative play, rigid thinking, and habitual behavior.
Exploration, the inverse of exploitation, involves taking risks. Creativity, mutation, and accident are all involved in patterns of exploration. Experimentation may lead a chess player to put knights in play before moving a pawn to E-4. Curiosity might lead a ground squirrel to approach a human being instead of running away. Spontaneous mutation may give a bacterium the ability to metabolize other kinds of sugar such as lactose. If any of these explorations meet with success-the chess player gets off to a good game by moving knights first, the squirrel gets a handout of nuts and seeds, or the bacterium encounters an abundance of an untapped resource-they will likely proliferate and be further explored and refined as established strategies.
But, statistically speaking, most explorations either lack attributable significance or simply fail outright. Our chess player could lose the game early. That squirrel could be chased off by a noisy child. A mutated bacterium might not encounter other types of sugar, gaining no advantage over its neighbors.
John Holland points out that there is an unfixed ratio between how much an individual explores versus how much it exploits in each agent within a system. How risk averse is the chess player? When does curiosity overcome fear in a small mammal? What is the genetic mutation rate as a bacterium reproduces? All of these qualitative questions point to this ratio between these general tendencies of exploitation and exploration. And this ratio can be extended and averaged statistically over a population, with some individuals more likely to explore than others.
According to Holland’s research, both traits are essential to a population’s survival. If a population exploits too much, it becomes rigid and brittle, easily wiped out by any environmental change. Likewise, if a population takes too many risks that add up to big failures, they too will disappear. For any iterated population to survive, be it a collection of strategies in a gamer’s brain, biologically biased behavior in a population of small mammals, or genetically determined properties, there must be a balance between exploitation and exploration. In most circumstances, this balance favors exploitation over exploration, but a complete absence of exploration spells certain death. By our natures as complex adaptive beings, each of us is a swirl of whimsy and reason. This is one duality to embrace!