LEARNING (Evolutionary reinforcement)
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(1) |
| ID | ◀ 1883 ▶ |
| Object type | General information |
A form of individual learning that should have an influence on the evolution of a group or a species.
This hypothesis was proposed by D.H ACKLEY (quoted by S. LEVY, 1992, p.265), who writes: “As dictated by their genetic code, agents would actually develop two neural networks: an ”action network“ that would convert sensory input into behaviors, and an ”evaluation network“ that would also draw from sensory input, by using that information to judge whether a particular situation was good or bad. Depending on the feedback it received after acting on that judgment, the agent would reinforce, or modify, its behavior” (1992, p.265).
It seems for example that individual ethological adaptation (e.g. the monkeys who learned to clean potatoes in sea salt water, in Japan), transmitted from generation to generation, modify the ecological niche of the species as a whole and could eventually lead to a different selection of random mutations.
ACKLEY successfully applied the idea to define a set of rules for evolving groups of artificial organisms, with positive results.
The model — a kind of Lamarckian transmission of acquired behavioral patterns — could probably be useful for the understanding of the way human organizations react to individual adaptations and innovations of some of their members.