I'm alone in my home, working in my study, when I hear the click of the front door, the sound of footsteps making their way toward me. Do I panic? That depends on what I — my attention instantaneously appropriated to the task and cogitating at high speed—infer as the best explanation for those sounds. My husband returning home, the house cleaners, a miscreant breaking and entering, the noises of our old building settling, a supernatural manifestation? Additional details could make any one of these explanations, excepting the last, the best explanation for the circumstances. Why not the last? As Charles Sanders Peirce, who first drew attention to this type of reasoning, pointed out: "Facts cannot be explained by a hypothesis more extraordinary than these facts themselves; and of various hypotheses the least extraordinary must be adopted."
"Inference to the best explanation" is ubiquitously pursued, which doesn't mean that it is ubiquitously pursued well. The phrase, coined by the Princeton philosopher Gilbert Harmon as a substitute for Peirce's term "abduction," should be in everybody's toolkit, if only because it forces one to think about what makes for a good explanation. There is that judgmental phrase, the best, sitting out in the open, shamelessly invoking standards. Not all explanations are created equal; some are objectively better than others. And the phrasealso highlights another important fact. The best means the one that wins out over the alternatives, of which there are always many. Evidence calling for an explanation summons a great plurality (in fact an infinity) of possible explanations, the great mass of which can be eliminated on the grounds of violating Peirce's maxim. We decide among the remainder using such criteria as: which is the simpler, which does less violence to established beliefs, which is less ad hoc, which explains the most, which is the loveliest. There are times when these criteria clash with one another. Inference to the best explanation is certainly not as rule-bound as logical deduction, nor even as enumerative induction, which takes us from observed cases of all a's being b's to the probability that unobserved cases of a's are also b's. But inference to the best explanation also gets us a great deal more than either deduction or enumerative induction does.
It's inference to the best explanation that gives science the power to expand our ontology, giving us reasons to believe in things that we can't directly observe, from sub-atomic particles — or maybe strings — to the dark matter and dark energy of cosmology. It's inference to the best explanation that allows us to know something of what it's like to be other people on the basis of their behavior. I see the hand drawing too near the fire and then quickly pull away, tears starting in the eyes while an impolite word is uttered, and I know something of what that person is feeling. It's on the basis of inference to the best explanation that I can learn things from what authorities say and write, my inferring that the best explanation for their doing so is that they believe what they say or write. (Sometimes that's not the best explanation.) In fact, I'd argue that my right to believe in a world outside of my own solipsistic universe, confined to the awful narrowness of my own immediate experience, is based on inference to the best explanation. What best explains the vivacity and predictability of some of my representations of material bodies, and not others, if not the hypothesis of actual material bodies? Inference to the best explanation defeats mental-sapping skepticism.
Many of our most rancorous scientific debates — say, over string theory or foundations of quantum mechanics — have been over which competing criteria for judging explanations the best ought to prevail. So, too, have debates that many of us have been having over scientific versus religious explanations. These debates could be sharpened by bringing to bear on them the rationality-steeped notion of inference to the best explanation, its invocation of the sorts of standards that make some explanations objectively better than others, beginning with Peirce's enjoiner that extraordinary hypotheses be ranked far away from the best.