Inference to the best explanation

DOI: 10.4324/9780415249126-P025-1
Version: v1,  Published online: 1998
Retrieved April 21, 2021, from

1. The legitimacy of inference to the best explanation

If inference to the best explanation is legitimate, then a person is entitled to accept a hypothesis provided it meets certain minimal standards in accounting for the relevant data, and explains the data better than any other hypothesis available. But, presumably, when we make an inference from a body of evidence, our goal is to arrive at a further or wider grasp of the truth. Inference to the best explanation will advance this goal only if the satisfaction of explanatory desiderata makes a hypothesis likelier to be true. The trouble is that explanatory virtues (like breadth or simplicity) and truth appear to be unrelated (see Theoretical (epistemic) virtues). Under these circumstances, believing a hypothesis because of its explanatory value would not be much better than believing it because someone thought it up on your birthday. It seems that such a procedure will, if anything, hinder the search for truth, not promote it.

Proponents of inference to the best explanation may respond to this line of criticism in a number of ways:

  1. They could deny that truth is the goal of inquiry (see Scientific realism and antirealism §§1–2). This reply may become more difficult to sustain when inference to the best explanation is used in everyday life. For example, it would be uncomfortable to think that a jury which rejects a contention on explanatory grounds (say, for having too many loose ends) is not seeking the truth.

  2. Another response would be to deny that the goal of inquiry is solely to amass a body of truths. Rather, we seek (true) explanations of phenomena. Explanatory goodness thus stands alongside likelihood of truth as a distinct but legitimate basis for evaluating hypotheses. This suggestion, however, does not seem to dispel the concern that there may be a conflict between pursuit of truth and pursuit of explanatory goodness.

  3. Let us say that a hypothesis is tested in so far as it, or its consequences, is compared with observed data, and let us grant that the successful testing of a hypothesis by observations (confirmation) increases the likelihood that the hypothesis is true. One might then try to argue that better explanations are more testable by a body of data than inferior ones. In that event, the superior explanation will be better confirmed by the data, and favouring the superior explanation would at least indirectly further the pursuit of truth.

Consider, for illustrative purposes, the following all-too-crude way of linking simplicity and testability. Suppose hypothesis H explains a given phenomenon by positing one mechanism A, while its competitor H* accounts for the same phenomenon by positing the joint action of two different mechanisms B and C. The simpler hypothesis H will be more readily tested because we need evidence to support only one independent claim (‘A is at work’), instead of two (‘B is at work’ and ‘C is at work’). A difficulty here is that one might just as well say that H is less testable than H*, since H* can be disconfirmed by finding evidence against either of its components. Friedman (1983) and others argue with much more sophistication and plausibility that: (a) explanatory hypotheses are ones that unify the data; and (b) to the extent that hypotheses participate in multifarious, unifying relations to the data, they are more thoroughly exposed to testing by that data. Hence, superior explanations are more strongly confirmed by observations, and have better claims to truth.

Citing this article:
Vogel, Jonathan. The legitimacy of inference to the best explanation. Inference to the best explanation, 1998, doi:10.4324/9780415249126-P025-1. Routledge Encyclopedia of Philosophy, Taylor and Francis,
Copyright © 1998-2021 Routledge.

Related Searches


Related Articles