System Identification of Decision-Making Process in Gold Trading Game

The decision-making process of human are being researched in association with making psychological models and experiment technique. However, there is no mathematical models using transfer functions for representing such process as far as the authors know. In this research, a model of the decision-making process was constructed, which was consisting of proportional, integrator, and derivative elements. And, the coefficients of the each element, Kp, Ki, and Kd, were individually identified by the least-square method, using the input-output data obtained in the computer simulation game of gold trading, which was created especially for this research.In the result, Ki/Kp, the coefficient ratio between the integrator element and the proportional element, and Kd/Kp, that between the derivative element and the proportional element, were different for each individual. As these differences can be regarded as individual differences, the constructed model is considered to achieve the representation of individual properties, which is the aim of this research.However, in this created game, the recall ratio of Ki/Kp and Kd/Kp was not high respectively, the average of them was only 15%. This low recall ratio can be due to the inadequacy of the game used for the identification, or the unsuitability of the structure of the constructed model. Thus, the improvement of the gold trading game and model's structure are the next challenges. Especially, the addition of learning function to the model of the decision-making process is one of the desirable challenges, because it could work even in short and repeated process like gold trading.