Inference to the best explanation

Lipton, P. (1991) Inference to the Best Explanation (London Routledge). 

In connection with Mendelian and molecular genetics, Kitcher (1984) called this "extended explanation" and Needham (1999b) in his discussion of Hanna s (1966) Quantum Mechanics in Chemistry, speaks of "overall coherence." It is one thing (while appealing to an inference to the best explanation) to say that a theoretical "picture" plays a unifying role, quite another to suggest that some isolated phenomenon can, in some sense, be derived from it. 

Unlike Kim, however, I do not think that the impossibility of either kind of functionalization entails that qualia are not physical properties. I think that if we were to find strict neuroscientific nomological correlates of qualia, then we would be justified in holding that qualia are identical with their neuroscientific correlates by an inference to the best explanation (McLaughlin, 2001, 2010, 2011). Kim (2005) has challenged the idea that identities can be inferred by inference to the best explanation for a response, see McLaughlin (2010, 2011). 

But I don t want to make too much of this linguistic fact, if it is a fact. Instead, we should be liberal, allowing both an opaque and a transparent sense of explain. So, in the transparent sense of explain, Kim is right, and in the opaque sense he is not. And that is enough for my point in one sense of explanation, the identities allow explanations one would not have without them. Is explanation in that sense enough to ground inference to the best explanation Yes. It is a fact that ice floats on water, and a view that allows an explanation of that fact — even if only an opaque explanation of it — is thereby made more reasonable to believe than views that do not allow such explanations. 

McLaughlin, B. P. (2010). Consciousness, type physicalism, and inference to the best explanation. Philosophical Issues, 20, 266—304. 

According to inference to the best explanation, hypotheses are supported by the very observations they are supposed to explain. Moreover, on this model, the observations snpport the hypothesis precisely because it would explain them. Inference to the BE thus partially inverts an otherwise natural view of the relationship between inference and explanation. According to that natural view, inference is prior to explanation. First the scientist must decide which hypotheses to accept then, when called upon to explain some observation, s/he will draw from her/his pool of accepted hypotheses. According to IBE, by contrast, it is only by asking how well various hypotheses would explain the available evidence that s/he can determine which hypotheses merit acceptance. In this sense, IBE has it that explanation is prior to inference [16]. 

Lipton P (2000) Inference to the best explanation. In Newton-Smith WH (ed) A companion to the philosophy of science. Blackwell, pp 184-193... 

Such an objection is valid. The CSIC experiment cannot categorically rule out the possibility that human intelligence is a kind of algorithm. However, CSIC can provide an inference to the best explanation between two competing hypotheses. The first hypothesis asserts that the proximate cause, human intelligence, created the information, while the second hypothesis states that a more remote intelligent agent created the information. 

The model of the inference of the best explanation is designed to give a partial account of many inductive inferences, both in science and in ordinary life. One version of the model was developed under the name abduction by Pierce [15] (early in this century and the model has been considerably developed and discussed over the last 25 years). Its governing idea is that explanatory considerations are a guide to inference that scientists infer from the available evidence to the hypothesis which would, if correct, best explain that evidence. 

This view, however, seems at odds with actual scientific practice. Actual scientific practice takes higher-level causal explanation and higher-level inductive inference to be some of the best explanations and inferences science can discover. 

An extended material of valence electron spatial correlations (VEC) had been analyzed (Schubert, 1964), when it became apparent that one correlation of valence electrons alone is not sufficient for the explanation of crystal structures of metallic phases. The outer core electrons had to be taken into consideration. This may best be seen from the crystal structure of indium (Fig. 4) The lattice matrix of In may be given in diagonal form ai = (4.59 4.59 4.95) A. The explicit lattice constants are needed for verification that the proposed VEC is acceptable. The VEC is aj = aAi(l, —1,0 1,1,0 0,0,3/2) and may be decomposed into the equations at = a j + a2 a2 = - aj + a2 a3 = 3 a3/2, which may be verified by means of Fig. 4. If a correlation lattice is inserted into a crystal structure, this does not mean that there are positions of increased electron density in the cell, it only gives the commensurability which is favorable energetically. It is easily verified that the number of valence electron places per cell is = 12 and is equal to the number of valence electrons in the cell given above = 12. The A1 type of the VEC had been inferred from the diamond struc-... 

It is clear that higher-level properties play an indispensable role in our inductive inferences. This supports the idea that the causal mechanism is also often at the higher level. Nonreductive materialism embraces this idea. It recognizes the fact that causation can occur at multiple levels of realization and that part of telling the best causal story is finding the appropriate level of causation and causal explanation. Good philosophy of science seems to tell us that causation can occur at multiple levels. Nonreductive materialism can subsume mental causation under an account of higher-level causation within the sciences. 

If the reductionist wants to remain a realist about causal explanation (as Kim does), then it seems that he will have to deny that our best causal explanations and our best inductive inferences can refer to higher-level properties. If all of these higher-level properties can be reduced in some fundamental sense to the lower level, then the best causal explanations we can offer and the best generalizations and inductive inferences we can make wiU refer to lower level properties. We can give explanations and make inductive inferences that refer to higher-level properties, but these will not be genuine explanations because they will not get the causal mechanism right. Our best explanations... 

Higher-level causation and higher-level causal explanation are indispensable parts of our lives. We have seen that the higher-level property of size helps us best understand and explain why the balls sort the way they do in Sober s well-known ball-sorter example. It is also the case that understanding that the balls sort in virtue of size will help us best predict how the balls will sort in the future. So causation, causal explanation, and inductive inference are all closely intertwined. The property in virtue of which an event is caused, the property to which we refer for the best causal explanation, and the property to which we look when making generalizations and inductive inferences will be often be one and the same property for any event we wish to understand. 

So why is it that some of our best causal explanations, generalizations, and inductive inferences often make reference to higher-level properties The obvious answer is that it is because these higher-level properties are causal properties. If this is right, then the reductionist must try to explain how it can be that some of our best causal explanations refer to higher-level properties if all real causation ultimately occurs at the lower level. 

---