Behaviorist stochastic modeling of instrumental learning.

Area of Science:

• Behavioral economics
• Mathematical psychology
• Reinforcement learning

Background:

• Instrumental learning, or operant conditioning, is a fundamental behavioral process.
• Understanding the dynamics of learning and decision-making is crucial in various fields.
• Existing models may not fully capture the continuous adjustment of behavior based on utility.

Purpose of the Study:

• To develop a novel mathematical model for instrumental learning.
• To describe the agent's learning process as a non-stationary Poisson process.
• To interpret the derivative of expected utility as the agent's drive for optimization.

Main Methods:

• Formulation of a mathematical model based on expected utility theory.
• Definition of utility as expected benefit minus expected cost.
• Derivation of an ordinary first-order differential equation governing act intensity changes.

Main Results:

• The model describes an agent learning to commit acts per time unit.
• The agent's drive is proportional to the derivative of expected utility.
• Act intensity changes are driven by this utility-driven "drive".

Conclusions:

• The presented differential equation offers a new framework for instrumental learning.
• The model provides insights into the continuous adaptation of behavior.
• This approach can be applied to understand decision-making under varying conditions.