Selection by reinforcement: A critical reappraisal.

Area of Science:

• Behavioral psychology
• Evolutionary biology
• Neuroscience

Background:

• The concept of ontogenetic selection by reinforcement posits an analogy between individual learning (conditioning) and species evolution (natural selection).
• This idea has been explored through Skinnerian (operant conditioning) and extended (including Pavlovian conditioning) frameworks, often linked to neural network models.

Purpose of the Study:

• To critically reappraise the concept of ontogenetic selection by reinforcement.
• To evaluate two main versions: the traditional Skinnerian approach and an extended version incorporating Pavlovian conditioning.
• To argue for the inadequacy and metaphysical unsoundness of selectionist accounts of conditioning.

Main Methods:

• Critical analysis of existing theoretical frameworks of ontogenetic selection by reinforcement.
• Examination of the explanatory power of these frameworks for both operant and Pavlovian conditioning.
• Evaluation of the necessity and utility of selectionist principles in conjunction with neural network models and genetic algorithms.

Main Results:

• Both Skinnerian and extended versions of ontogenetic selection by reinforcement are unable to adequately capture Pavlovian conditioning.
• The concept of selection by reinforcement is deemed unnecessary for formulating and utilizing neural network models for explanation and prediction.
• Non-selectionist accounts of conditioning are presented as more intelligible, explanatory, and heuristic.

Conclusions:

• The analogy between individual learning and phylogenetic evolution via selection by reinforcement is flawed.
• Alternative, non-selectionist explanations of conditioning offer superior scientific utility.
• The metaphysical underpinnings of selectionist learning theories are questionable.