Modeling Gambling: An Application of the Mathematical Principles of Reinforcement

Abstract

The Mathematical Principles of Reinforcement (MPR) has proved a useful model for predicting and describing the behaviour of non-human animals on different schedules of reinforcement. This research tests the ability of MPR to accurately predict performance of adult humans on a simulated gambling task. A simulated electronic gaming machine was used in three experiments and gambling responses were reinforced according to series of Random Ratio schedules. In Experiment 1, when participants experienced either an ascending or descending order of ratios, rates of responding were well described by a bitonic response gradient. In Experiments 2 and 3 participants experienced either an early large win or an early large loss before experiencing a series of ratio schedule values that were presented in ascending order. Again rates of responding, expressed as a function of ratio schedule value, were well described by a bitonic response gradient. The early large loss condition produced higher response rates than the early large win condition. The bitonic response gradients of all conditions were well described by MPR via changes in the parameter a, specific activation.