Network and direct methods of maximising harmony

Paul Thagard (1989t) presents an interesting and important theoretical account of choosing between hypotheses in scientific and everyday domains, describing a network model, ECHO, which aims to embody these ideas. Thagard's integration of philosophical, psychological, and computational considerations is impressive in its breadth, theoretical rigor, and conceptual simplicity. It is a compelling example of the value of interdisciplinary research in the understanding of scientific and everyday reasoning. The general problem of choosing between alternative hypotheses, explanations, or theories is so profound that it might be thought that modeling is wholly infeasible. Thagard is able to finesse nicely many thorny problems, however (by presupposing a formalisation of the domain of application into discrete propositions representing hypotheses and evidence, and taking the degree of coherence between these propositions as given), leaving a constrained yet important subproblem to be solved. Central to Thagard's thesis is the claim that theory choice can be based on the "explanatory coherence" of sets of propositions, where "the global explanatory coherence of a system S of propositions is a function of the pairwise local coherence of those propositions" (Thagard 1989t, p. 437). In theory choice, the set of propositions with the maximal explanatory coherence is preferred. Formally, Thagard measures explanatory coherence as a simple function if: