Learning curve for collective behavior of zero-intelligence agents in successive job-hunting processes with a diversity of Jaynes-Shannon's MaxEnt principle

摘要

Inspired by the unsupervised learning or self-organization in the machinelearning context, here we attempt to draw `learning curve' for the collectivebehavior of job-seeking `zero-intelligence' labors in successive job-huntingprocesses. Our labor market is supposed to be opened especially for universitygraduates in Japan, where the students have several successive chancesn=0,1,2,... to obtain their positions within an academic (business) year. Inthis sense, the `cumulative unemployment rate' in our model system is regardedas an error-measurement in the collective intelligence of students, and thejob-hunting stage n-dependence of the error constructs a learning curve. In oursimple toy-model of probabilistic labor market, the diversity of students'behavior is built-in by means of the Jaynes-Shannon's MaxEnt (Maximum Entropy)principle. Then, we discuss the speed of convergence for the error-measurement,where we consider a scenario in which the students do not use any informationabout the result of job-hunting processes in the previous stage. Our approachenables us to examine the existence of the condition on which macroscopicquantity, say, `stage-wise unemployment rate' becomes `scale-invariant' in thesense that it does not depend on the job-hunting stage n. From the macroscopicview point, the problem could be regarded as a human resource allocation.