The aim of this paper is conceptual and theoretical: to propose the concept of generic scientific task as an organizing principle for research on scientific discovery, or more broadly, scientific inference. Generic tasks cut across various scientific fields, but involve rather specific inferences; in any case, their generality is found only by surveying science as a whole. Computer programs that carry out these generic tasks may differ somewhat from one science to another, but they all share similar computational mechanisms: hence they represent generic tasks. We assemble a broad array of previous work as evidence to support this concept, and work out the research plan and methods that are implied by it.

full paper