In Response to Craver: A Defense and Critique of the HPC View of Natural Kinds (2)

Continued from page 1

 

Objections

Let us now turn to Craver’s objections to the HPC view and his contention that it does not avoid incorporating conventionalist elements into its theory.  The first challenge Craver poses for the HPC view is to specify which homeostatic mechanisms are identified with which natural kinds and to give an explanation for this association that does not rest on perspectival or pragmatic considerations (Craver 578).  Craver points out that any given natural kind is involved in many homeostatic mechanisms and that “one can be led to lump or split the same putative kind in different ways depending on which mechanism one consults in accommodating the taxonomy to the mechanistic structure of the world” (Craver 583).  That is, if we attend to some mechanisms, we will lump our kinds; if we attend to others, we will split our kinds (Craver 584).  Thus, it appears that there is no final and correct way of delineating the kinds objectively and naturally.

If this is the case, then what ultimately decides whether we should lump kinds or split a kind?  Craver believes that pragmatic and perspectival considerations decide.  We should be pluralists about kinds, taking there to be “as many kinds as there are distinct and dissociable mechanistic entanglements,” and when conflict arises as to how we should define a kind, we should make our decision based upon what we are trying to “understand or to do” (Craver 584).  This means that two scientists that are trying to solve different problems may legitimately classify kinds in different ways to suit their purposes, and neither will be able to claim that her taxonomy is more “correct.”  While both will rightly appeal to the causal structure of the world to ground their claims, they will be attending to different mechanisms to define the kinds, and there is no way of privileging one classification scheme over another except by appealing to pragmatic considerations.  But as soon as we appeal to pragmatic considerations in our taxonomy, we will have introduced a conventionalist element into our taxonomy and the HPC theorist can no longer say that the HPC view delivers kind classifications that are wholly objective and natural.

Craver’s second challenge to the HPC view is that it cannot tell us when two mechanisms are of the same kind.  Identity conditions of mechanisms need to allow for “variability among instances of the kind” and allow for “variability in the mechanisms” that realize the kind (Craver 586).[3]  Given this variability, we have a situation that is similar to the Generality Problem for process reliabilism in epistemology, namely, that we cannot specify the appropriate level of generality for the kinds of mechanisms with which we wish to identify kinds.  Craver claims that too abstract a characterization will gloss over sub-kinds in mechanisms while too detailed a characterization will “make each particular mechanism a kind unto itself” (Craver 587).  Between these two extremes there are no objective features to ground a certain level of abstraction, and so what will decide the appropriate balance of abstraction and detail in the characterization of a kind of mechanism will be “what we want to do with the schema” (Craver 588).  That is, practical and theoretical considerations will decide how we characterize a kind of mechanism, and not the objective features of the natural world.  And “if there is no objectively appropriate degree of abstraction for typing mechanisms, then judgments about whether two mechanisms are mechanisms of the same kind rely ineliminably on judgments by people (in concert) about the appropriate degree of abstraction required for the problem at hand” (Craver 589).  To allow for this is to allow for conventionalist elements to creep back into the HPC view of natural kinds and to lose the purely objective feature that HPC kinds are supposed to have.

The third challenge Craver poses to the HPC view of kinds is that it cannot provide objectively determined boundaries for mechanisms; it cannot tell us which “entities, activities, and organizational features are part of a mechanism (or kind of mechanism) and which are not” (Craver 589).  In the absence of purpose, goals, forms, designs, and teleology in nature, Craver claims that we cannot divide entities into parts (Craver 589).  And since we cannot say which parts are involved in an entity, we cannot specify the boundaries of that entity.  As Craver puts it, “the spatial and causal boundaries of mechanisms depend on the epistemologically prior delineation of relevance boundaries. But relevance to what?  The answer is: relevance to the phenomena that we seek to predict, explain, and control” (Craver 590, original emphasis). 

Consequently, the boundaries of mechanisms will be fixed by pragmatic and theoretical considerations of scientists and not solely by the objective and natural features of mechanisms.  Mechanisms will contain “all and only the entities, activities, and organizational features relevant to the phenomenon selected as our explanatory, predictive, or instrumental focus” (Craver 590).  If this is the case, then how we explain the structure of the natural world will depend “upon our explanatory interests and our descriptive choices at the level of property clusters,” meaning that conventionalist elements have once again entered into our supposedly objective classificatory scheme of the natural world (Craver 590-1).  Thus, HPC natural kinds are not truly natural in the full sense that HPC view adherents are supposed to desire.

A fourth concern that Craver has about the HPC view is that scientists define kinds by property clusters that figure in “important causal generalizations” (Craver 578).  Craver focuses on the word “important” because this rules out defining kinds according to property clusters sustained by mechanisms that “are of little or no theoretical or practical value” (Craver 578).  As soon as we include normative notions such as “important” into our classification scheme, we will have introduced a conventionalist element since what is “important” depends on “whether the kind in question appears in our theories or is otherwise important for our aims and objectives,” and this is to destroy the purely objective and natural divisions of kinds (Craver 579).  One might give up on “important” causal generalizations and say that kinds are grouped by a “mere causal relevance” of the cluster to something else.  However, Craver argues that this solution increases promiscuity such that “there are many more natural kinds than science will ever find it useful or interesting to recognize, let alone study,” and presumably, this is an unacceptable result (Craver 579).

Fifth and finally, Craver takes the Accommodation Thesis to be flawed.  Craver contends that both readings lead to problems in that the strong reading is “untenable” while the weak reading is “too weak to avoid conventionalism” (Craver 580).  The strong reading is untenable because our kinds are revised upon further evidence that a differing scheme better fits with the causal structure of the world, and there is no guarantee that the kinds that we use now will in fact be used in the future (e.g., phlogiston).  The weak reading is also problematic because things that are “similar enough” will use terms that gloss over causal differences for the sake of defining a kind that fits human concerns, and this is to make kinds subjective and conventionalist.

Continued on page 3