Power Law of Practice in Adaptive Training Applications.

摘要

This research paper was inspired by the difference of findings in empirical learning-curve-data collected by ourselves and those of Newell and Rosenbloom (1981), who analysed empirical data-sets from other investigators to prove the ubiquity of the Power Law of Practice. The aim of this paper is to show some fundamental statistical and numerical flaws associated with Power Law analysis in this research literature. Additionally, we present the results from analysis of empirical data that reveals the low predictive value of the Power Law of Practice, using leastsquares non-linear regression technique to find the best fitting estimate for the free parameters of the Power Law. These exercises lead us to the recommendation to take account of these flaws, for example, when considering application of the Power Law of Practice in adaptive training or intelligent tutoring.