On Classification in the Study of Failure, and a Challenge to Classifiers

Classification schemes are abundant in the literature of failure. They serve a number of purposes, some more successfully than others. We examine several classification schemes constructed for various purposes relating to failure and its investigation, and discuss their values and limits. The analysis results in a continuum of uses for classification schemes, that suggests that the value of certain properties of these schemes is dependent on the goals a classification is designed to forward. The contrast in the value of different properties for different uses highlights a particular shortcoming: we argue that while humans are good at developing one kind of scheme: dynamic, flexible classifications used for exploratory purposes, we are not so good at developing another: static, rigid classifications used to trap and organize data for specific analytic goals. Our lack of strong foundation in developing valid instantiations of the latter impedes progress toward a number of investigative goals. This shortcoming and its consequences pose a challenge to researchers in the study of failure: to develop new methods for constructing and validating static classification schemes of demonstrable value in promoting the goals of investigations. We note current productive activity in this area, and outline foundations for more.