The problem of induction is to validate inferences from some experiences or data to others. By experiences or data I mean such things as red patches, sounds, tastes, pains. Distinctions between private and public data, between internal and external impressions, between data and objects, are not epistemologically primitive or given but are modes of categorizing the given. The application of categories and the construction of objects are cases of, and so presuppose the validity of, induction.
To hypostatize a construction, or a system of constructions, is in fact to generate what may be called an external world. Examples are physical objects, and scientific objects. And when analysis proceeds, as it often does, by replacing the original subject by constructs, a hierarchy of external systems is generated. Thus, physical objects—e.g., our bodies—are external to data, and scientific objects are external to physical objects.