Analysing recovery from pandemics by Learning Theory: the case of CoVid-19

Romney B. Duffey; Enrico Zio

Learning theory has been successfully applied to quantify error and learning rates in technical systems, casualties in large land battles, everyday accident and event data, and to human, software and hardware reliability (Duffey and Saull, 2008; Duffey & Ha, 2010; Fiondella & Duffey, 2015; Duffey, 2017a) . The novel feature is to replace calendar time or test interval, which has always been used before, with a measure for the accumulated experience and/or risk exposure, thus defining rate trends and quantifying effectiveness of responses to errors and accidents, and allowing totally different systems to be directly intercompared. Additionally, the trend is governed by two parameters that are physically based: the learning rate constant and the minimum achievable error rate. This is in contrast with statistical analysis ,where fitting to learning data is typically done on three empirical parameters (Bush and Mosteller, 1955 ) , and with the inverse "power laws" extensively fitted in cognitive psychology data ( e.g. Anderson, 1990 and the references therein).