3.3 Sample A – Cases of “Explain”
3.3.12 Preliminary Evidence for the Account
Having normalized all of the cases in sample A we can consider all the pairs of the types of the explanantia and explananda. While any of the 25 pairings is possible
prima facie, we see that not all of them occur in the data set. Table 3.2 shows the results, and figure 3.3 presents a visualization.
In sample A all of the explanations fall on the left-lower half of the table. There are model-kind explanations, but no kind-model explanations. There are theory- data explanations but no data-theory explanations. There are also some kind-kind explanations. The two most common forms are model-kind and model-data.
Given the prevalence of model-kind and model-data explanations, we might be tempted to say that when scientists say “explain” they mean that they have a model that explains a kind or some data about entities of that kind. I allow that model-kind explanation is the prototypical use of “explain”. But while model-kind explanation is more common than the other forms, by focusing on models as the explanans we ignore half of the explanations in sample A. I have argued that there is a similar structure in each case, with the key difference being the core relation. In the next section I will defend the claim that the core relations for all the forms share a common thread: they all support the counterfactual reasoning that is required for explanation.
This evidence provides some preliminary support for my account: it is general enough to cover a wide range of cases, and any additional complexity is justified by the flexibility to cover many forms of explanation that occur in sample A. Since the account was designed specifically to handle the cases in sample A, this is no surprise. The account must be tested against a wider range of cases to see whether it stands up.
data en tit y kind mo del theory data 0 0 0 0 0 entity 2 3 0 0 0 kind 1 1 3 0 0 model 5 1 9 0 0 theory 2 1 1 0 0
Table 3.2: The number of cases in sample A for each of the 25 forms of explanation, where each form is the pair of an explanans type (row) and an explanandum type (column), and the five categories of data, entity, kind, model, and theory classify both explanans and explanandum.
data
entity
kind
model
theory
data
entity
kind
model
theory
3.4
A General Philosophical Account of Scientific
Explanation
In scientific explanation the explain-relation holds between two qualities: the ex- planans quality and the explanandum quality. The explanandum quality depends on the explanans quality. If the explanans quality were different then the explanan- dum quality might also be different. Explanations allow us to reason counterfactually about the dependence of the explanandum on the explanans.
This much has been said before by philosophers of science. Woodward’s influ- ential account of explanation in terms of ideal interventions in causal processes is a counterfactual model of explanation (Woodward 2003). And the same insight can also be seen in other philosophical accounts, such as mechanistic, asymptotic, and model-based explanations. But the account I am proposing allows us to go further.
The explain-relation is always layered on top of some other relation between the explanans and the explanandum. I call this the “core relation”. What I have called the “form” of the explanation determines, in a general way, the nature of the core relation. For example, in a model-kind explanation such as A14 the core relation is modelling. The counterfactual dependence of the eventual occurrence of monoclonal conversion on the balance of the population holds only to the extent that the model of the reproduction of the equipotent stem cells models crypts of stem cells in the maintenance of intestinal epithelial tissue. In a kind-data explanation such as example A12 the core relation is a composite of the instantiation of the kind by the entity (i.e. GNAIW is a water mass) and the measurement relation between the entity and the data (i.e. between GNAIW and the measurements of δ13C in end-members). If GNAIW were not a water mass, the explain-relation would fail. If the measurements were not of GNAIW end-members, the explain-relation would fail to hold.
In general, the counterfactual supporting explain-relation between the explanans quality and the explanandum quality will hold as long and to the extent that the core relation holds between the explanans and the explanandum. We understand the details of the explain-relation in a given case by understanding the details of the counterfactual reasoning that the core relation supports.
Although the core relations can vary widely in different explanations across the range of scientific disciplines, the explain-relation is essentially the same. Scientists use “explain” in many different contexts, asserting an explain-relation between the qualities of various sorts of explanantia and explananda. The diversity of the uses of “explain” makes sense once we understand that there are diverse core relations sup-
porting the explain-relation in different cases. Scientific disciplines may rely on some forms of explanation more than others. The five coarse categories I have proposed, and the forms of explanation based on them, are only a starting point, and there is room for more detailed analyses of specialized forms of explanation that may be unique to particular scientific disciplines.