Thinking and Its Descriptions: Conceptions of Inference

To make a prediction or generalization based on descriptions of evidence is to accomplish an inductive inference. We constantly make such inferences, but we when we try to give reasons to defend our conclusions, it is not always clear what counts as a good reason and why. What is the difference between good and bad inductions? The difficulties philosophers have encountered in attempting to answer this question has led to an even more radical problem, first articulated by Hume: what justifies induction as a form of reasoning in the first place? Is there such a thing as a good induction? In recent literature, contemporary philosophers often proceed as if the Humean problem has been solved or is irrelevant. In doing so, I argue, they operate with an impoverished and overly narrow conception of inference. Inductive reasoning is often reduced to procedures for tracking and measuring pre-articulated, antecedently given facts. I argue that only once we have enriched our conception of reasoning will we be in a position properly to defuse the Humean challenge. To this end, I pursue an alternative philosophical approach to induction according to which inductive reasoning is not reasoning about correlations, and it is not a game played only once we’ve been provided with pre-articulated facts. Rather it is essentially concerned with how to describe phenomena and the kinds relevant to their proper description: First, the validity of inductive reasoning is radically context-dependent; it depends on our rational entitlement to forms of description in particular contexts. Second, key cases of inductive reasoning depend upon conceptual innovation—on ways in which the reasoning itself develops, changes and expands our concepts. This allows for a new form of solution to the problem of induction according to which inductive reasoning depends on analogical reasoning.