ABSTRACT
Diagnostic reasoning consists in going back to the causes that triggered one or multiple effects. Formally, it means estimating the probability Pr(cause|effect). It is well known that the estimation of this probability deviates from the rational norm defined by the Bayes’s rule. Recent results show that people follow very varied strategies in order to estimate the diagnostic probability. Most often, the encountered strategies consist in combining the quantities within the Bayes’s rule in a sub-optimal fashion. In this chapter, it is suggested that the existence of additional strategies is based on patterns of defeasible reasoning. In particular, two schemes appear to be psychologically plausible. The first is a form of defeasible deduction based on the Modus Ponens (DMP), the second is a scheme of defeasible abduction based on Affirming the Consequent (DAC). In an experiment, it is shown that the choice of the scheme to estimate the diagnostic probability Pr(cause|effect) depends on the value of the predictive probability Pr(effect|cause). When Pr(cause|effect) < Pr(effect|cause), participants do not report any particular preference between DMP and DAC. Yet, when Pr(cause|effect) > Pr(effect|cause), participants report a clear preference for DMP to estimate the diagnostic probability.
