Explanatory inference under uncertainty

Abstract

This paper investigates the performance of explanatory or abductive inference in certain hypothesis selection tasks. The strategy is to use various measures of explanatory power to compare competing hypotheses and then make an inference to the best explanation. Computer simulations are used to compare the accuracy of such approaches with a standard approach when uncertainty is present and when several causal scenarios occur including one where the conditions for explaining away are met. Results show that some explanatory approaches can perform well and in certain scenarios they perform much better than the standard approach. © 2014 Springer International Publishing Switzerland.