Uncertainty and imprecision play a leading role in human reasoning, therefore it is worth devoting some effort to the study of reasoning methods under uncertain data. On the other hand, abductive reasoning is an important component of many tasks, diagnosis among them.

This papers deals with an application of abductive reasoning to the handling of incomplete information; its main goal is to characterize the best decision in the framework of a rule-based system on hypothesis driven reasoning.

Technically, the focus is on fuzzy gradual rules, which are represented by residual implications, and in obtaining fuzzy abductive inferences based on fuzzy (or crisp) observations. A coherence measure is defined, which enables the reader to evaluate hypotheses and select the most coherent one.

A number of hypotheses are generated by abductive inference, and then measures of similitude are used in a second step to refine abductive explanations. As a result, a new approach to decision making is obtained by a clever mixture of two well-known tools, such as the abductive reasoning and the use of similitude relations.