Book Review: Inference to the Best Explanation by Peter Lipton

Inference to the Best Explanation (2nd  Edition) by Peter Lipton  is a 200-page philosophy text dealing with the process of how we come to conclusions from the evidence available to us. The book is technical, yet readable. The first three chapters provide an initial survey of the problems of induction and explanation. The middle section explores inference to the best explanation, methods of induction, and compares other models of explanation. The final section deals with some problems and common objections to the inference to the best explanation (IBE) model. This review will provide a general overview of some of the key ideas presented in the text.

Lipton discusses the ins and outs of induction and explanation, pointing out that proof is elusive and we generally work with probabilities: “Inductive inference is thus a matter of weighing evidence and judging probability, not of proof.” (284) Furthermore, “strong inductive arguments are those whose conclusions predict the continuation of a pattern described in the premises.” (380) With a nod to Hume, Lipton points out the problem with circularity in the inductive method: “The trouble is that the argument that conservative inductions will work because they have worked is itself an induction.” (394) Lipton also notes: “It is a striking fact about our inductive practice, both lay and scientific, that so many of our inferences depend on inferring from effects to their probable causes.” (585)

After laying some groundwork as well as exploring the method of agreement and the method of difference, the author delves into the inference to the best explanation model. “According to Inference to the Best Explanation, we infer what would, if true, be the best explanation of our evidence. On this view, explanatory considerations are a guide to inference.” (625) This model focuses on how our inferences can bring us to understanding: “The question about explanation can then be put this way: What has to be added to knowledge to yield understanding?” (637) The IBE model inevitably runs into the issue of infinite regression: “But if only something that is itself understood can be used to explain, and understanding only comes through being explained by something else, then the infinite chain of whys makes explanation impossible.” (646) Lipton notes that a stopping point is needed. “Sooner or later, we get back to something unexplained, which ruins all the attempts to explain that are built upon it.” (648) Lipton’s solution shows that we don’t need to know all the answers to know some of the answers: “A better reply is that explanations need not themselves be understood.” (651) The author spells out just what the infinite regress issue means to the IBE model:

Rather than show that explanation is impossible, the regress argument brings out the important facts that explanations can be chained and that what explains need not itself be understood, and so provides useful constraints on a proper account of the nature of understanding. (655)   

Lipton continues with discussion and comparison of various models of explanation, showing how they differ from IBE. “According to the causal model of explanation, to explain a phenomenon is simply to give information about its causal history.” (830) Inference to the best explanation looks at the data and our background beliefs, then “we infer what would, if true, provide the best of the competing explanations we can generate of those data (so long as the best is good enough for us to make any inference at all).” (1415) Narrowing down the “live options” is key in the model: “We have to produce a pool of potential explanations, from which we infer the best one.” (1451)

When using IBE, one need only look for a “potential explanation: there is no requirement that the explanation be true, only that it include a general hypothesis and entail the phenomenon.” (1454) Lipton makes additional points to ensure clarity along the way: “According to Inference to the Best Explanation, then, we do not infer the best actual explanation; rather we infer that the best of the available potential explanations is an actual explanation.” (1458) In addition, Lipton makes a key differentiation between probable explanations and “lovely” explanations, noting that the most likely is the one with the most warrant, yet: “On the other hand, we may characterize the best explanation as the one which would, if correct, be the most explanatory or provide the most understanding: the ‘loveliest’ explanation.” (1483) He spells it out more succinctly, noting that: “Likeliness speaks of truth; loveliness of potential understanding.” (1486) He notes the reason why these two elements don’t exactly overlap: “One of the reasons likeliness and loveliness sometimes diverge is that likeliness is relative to the total available evidence, while loveliness is not, or at least not in the same way.” (1493)

Lipton shows that it is not enough simply to look for the likeliest cause. In IBE, one looks for understanding. Lipton notes that “for Inference to the Best Explanation to provide an illuminating account, it must say more than that we infer the likeliest cause.” (1509) Furthermore, “Inference to the Best Explanation is an advance only if it reveals more about inference than that it is often inference to the likeliest cause. It should show how judgments of likeliness are determined, at least in part, by explanatory considerations.” (1511) The author stresses the need to assess an explanation based upon the illumination or understanding that it brings if it were actually true “Inference to the Best Explanation also suggests that we assess candidate inferences by asking a subjunctive question: we ask how good the explanation would be, if it were true.” (1605)

Some of the other abductive approaches that Lipton assesses, at least briefly, include Bayesian probability; as he notes, “the Bayesian and the explanationist should be friends.” (2846) But before moving on to look at some difficulties in the IBE model, Lipton summarizes the overall approach nicely:

First we identify both the inferential and the explanatory virtues. We specify what increases the probability of a hypothesis and what makes it a better potential explanation; that is, what makes a hypothesis likelier and what makes it lovelier. Second, we show that these virtues match: that the lovelier explanation is the likelier explanation, and vice versa. Third, we show that loveliness is the inquirer's guide to likeliness, that we judge the probability of a hypothesis on the basis of how good an explanation it would provide. (2850)

One of the objections raised against IBE comes from Voltaire, who asks why we should even think that the best explanation is true: “But why should we believe that we inhabit the loveliest of all possible worlds? If loveliness is subjective, it is no guide to inference; and even if it is objective, why should it line up with truth?” (3336) This does bring one back to presuppositions, or background beliefs, as Lipton notes: “inference also depends strongly on background beliefs, and these too will vary from person to person.” (3340) But the author notes that despite such objections, “Inference to the Best Explanation is in no worse shape than any other account of induction.” (3434)

In sum, Lipton’s Inference to the Best Explanation is a helpful read for those concerned with weighing evidence, assessing conclusions, and even evaluating worldviews. Lipton’s book certainly contains much more depth than this brief overview can show. For those interested in epistemology or philosophy of science, this is a worthy read.

Peter Lipton, Inference to the Best Explanation (New York, NY: Routledge, 2004), Amazon Kindle edition.