REINFORCEMENT
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(2) |
| ID | ◀ 2813 ▶ |
| Object type | General information |
The strengthening of some characteristic in a system by repetition of a stimulus-response effect.
Reinforcement cannot occur against unfavorable selection pressure. It generally starts as a random effect, which leads to a more satisfactory situation and thus becomes favored if repeated.
Reinforcement can be automatic, as for example in natural selection resulting in the multiplication of some genes in a population after a mutation, favorable in specific conditions. It has also been considered as a basic condition of animal or human learning by behaviorists. Devoid of its purely mechanistic taint, it is not rejected by so-called cognitivists, provided it be conjugated with recognition memory. (H.L. PETRI and M. MISHKIN, 1994, p.30-37).
In learning, according to these authors: “The role of reinforcement… is not to change the probability of a given response, but to lead to the development of expectancies about how certain goals can be reached” (Ibid., p.30). Reinforcement can possibly be used as a tool by purpose seeking systems.
The concept also can be extended to recognition processes in automata and, possibly to global behavior in automata networks.
See also
Hebb's rule