The Case for Boyd’s Homeostatic Property Clusters

clustet342What makes a bird a bird? Well, our taxonomy of course. What if this changes though over time and evolution? What if the properties we associate with birds one day no longer exist or only some? Boyd offers an account of classifying natural kinds, such as birds or psychiatric disorders based on properties which usually cluster together because of causal relations between them. I try to defend that view here. After thoroughly setting up the HPC theory in contrast to essentialism, with particular emphasis on its homeostatic causal mechanism, I shall then try to defend this mechanism against essentialist attack. I will examine how clustering is contingent on the causal structure of the world with examples from biological taxa and psychiatry.

Questions about natural kinds concern our ability to classify entities in nature, such as elements, species or social kinds. The HPC theory of natural kinds stands opposed to “traditional views” of kind meaning wherein semantics are decided on the basis of property-conjunction (Putnam, 1975: 140). A kind has a set of necessary and sufficient properties or “essence” that if met entail membership of an entity. A defender of this view is Ellis who asserts that “membership of a natural kind is decided by nature, not us”, or that these divisions naturally exist and ought to be discovered (2001: 19). Problematically, it is not the case that all kinds have sufficient properties without exceptions leading to empirical issues in kind classifications, e.g. in biological taxa (Wilkerson, 1995). A new account was needed to aid classification, especially in the special sciences. The motivation for HPCs is hence less semantic than epistemological in that they attempt to give guidance to scientists in classifying kinds to make inferences and explanations that “contribute to our epistemic reliability” (Boyd, 1999: 146). An advantage of HPC theories which distinguish kinds in terms of sets of properties that not all members have to share is that they do not require finding a “core set” of properties for all members to correspond to (Reydon, 2009: 6). This flexibility is difficult to conceptualise and the attractiveness of the cluster intuition seems inversely correlated to philosophers’ ability to abstract essences (Slater and Borghini, 2011: 4).

This is where Boyd comes in, whose HPC theory is currently one of the most popular accounts for natural kindhood (Reydon, 2009: 2-3). It asks us to be realist about kinds since they are indispensable in scientific and epistemic practices (Slater and Borghini, 2011: 2). At the same time it addresses that many kinds are not sets of identical entities, but instead are groups of entities bearing “various degrees of causally supported resemblance”, meaning that they possess significantly similar properties as a result of the same causes of those properties (Boyd, 1999: 144). This means that there are properties which are found to “cluster together in natural entities” sustained by some causal mechanism (Reydon, 2009: 6). Note that the clustering of properties without a mechanism is not sufficient for classification since if properties alone could define a kind then essentialism is true. Instead HPC kinds are classified in terms of “mutually reinforcing networks of causal mechanisms” (Kendler, et al., 2011: 1143).

What could this homeostatic mechanism be? In order to account for both the “non-accidentality” as well as “imperfection” of clusters (Slater and Borghini, 2011: 5), Boyd needs to tackle how clusters “reside together” and are “stable” in their unity (Kornblith, 1993: 7). He does this less conceptually as the essentialist than causally. To give an example: Boyd’s primary illustrations are biological taxa (see exemplary 1991, 1999, 2010). Here the members of a species usually exhibit some properties such as size and habit. The homeostatic mechanism that clusters these are for instance shared ancestry and “reproductive cohesion” (Boyd, 1991: 167), all of which can change through evolution. It is the “variation of genes and environment that produces different phenotypes” which makes the quest of finding necessary and sufficient properties an “idealisation” (Boyd, 1999; Wilson, et al., 2007). The HPC theory also accurately predicts sub-populations, genetic divides, hybrids and extraordinary behaviours. According to Boyd, we ought to imagine a matrix containing all properties of organisms, including genetics and ecology. Clusters of organisms would be comparably stable in this matrix. In such locations of clusters we find a fuzzy bordered species dependent on “homeostatically related morphological, physiological and behavioural features” (Boyd, 1991: 142). However, not all traits need to overlap and the important point to make is that the species borders’ “fuzziness … does not detract from their stability” (Kendler, et al., 2011: 1147).

Under this light essentialism does badly in describing biological phenomena. Its problem stems from the necessary variation among members of a species. A core set of sufficient properties would not make sense insofar as there is no reason to maintain that a species will always carry on some salient features without change. In response, Devitt believes that the ability to make generalisations about species, mainly genetic, must be indicative of an essence (2008: 352). I believe this is false. There is property clustering in the world: Our ability to make generalisations stems from shared properties of members, “to whatever degree” (Chakravarrty, 2007: 170) as in stem cells, which possess a number of characteristics with exceptions, such that they represent “a genuine HPC kind” (Wilson, Barker and Brigandt, 2007: 218).

One example in particular that I would like to examine are psychiatric kinds. They too are argued to be real  by practitioners since they are “objectively grounded features of the causal structure of the mind/brain” and are stable across cultures and time independent of human classification (Kendler, et al., 2011: 1143, 1147). The stability of symptoms, for instance schizophrenia is maintained by causal mechanisms between physiological and computational levels, where no one level can capture the whole maintenance (Kendler, 2008; Craver, 2007 on levels). Symptoms also are not grounded in a common cause for each individual but instead stand in “direct causal relations to each other” (Borsboom, 2008: 1101), multiple mechanisms from evolution, development or experience contribute to the expression of a kind and no causal mechanism has entitlement over another (Kendler, 2008). For example, self-enforcing symptoms, as can be found in depression (see Beck and Alford, 2008) can best be regarded as residing in a causal structure influencing each other which ensures their continued co-occurrence. The psychiatrist’s task is finding these complex “multi-level causal mechanisms that produce, underlie and sustain psychiatric disorders” instead of searching for core properties of disorders, which is exactly the move Boyd tries to make (Kendler, et al., 2011: 1146). In essence, scientific and medical practice just uses HPC theory kinds “regardless of whether they fit into the essentialist picture” (Reydon, 2009: 5).

This flexibility however yields room for essentialist critique: After identifying all members of a kind, their properties and homeostatic mechanisms, there is still no way to determine the kind’s extension since properties and mechanisms are allowed to change over time. This makes drawing boundaries between species conceptually impossible. The essentialist can assert that what follows is that HPC theory is scientifically useless in determining criteria for kind membership by which to judge real entities. Boyd himself points this out but as a reminder that HPC theory is realist. “Indeterminacy” of an extension is a necessary element of HPC kinds as it reflects the actual state of affairs in nature, which are “vague around the edges” (1991: 141). I am also inclined to dismiss this critique on the grounds that it misunderstands Boyd’s goal, which is not to determine the extensions of kinds, but to contribute to epistemic reliability of making inferences using kinds in practice as in psychiatry. (This is achieved with Boyd’s semantics of the accommodation thesis whereby causal mechanisms guide scientists in accommodating their categories to clusters to make more successful inductions; for discussion see 1991: 139; 1999: 147).

There are many open questions the essentialist can pose in developing a homeostatic causal mechanism. The very name suggests its requiring some understanding of causation, yet Boyd blatantly ignores this problem in his papers, while I believe it is very plausible that the nuances of whatever account for causation we chose will have great impact on the HPC-theory. Strevens would disagree with me, suggesting that we ought to make only few presumptions on causation when engaging with the HPC-mechanism since it is not the nature of causation, but the nature of epistemic reliability that is important (2008: 5; in Slater and Borghini, 2011: 10). Still, there are a few presumptions and how they should be accounted for is unclear. Quintessentially, although none of the essentialist’s counterarguments to a causal mechanism seems prima facie convincing, Boyd’s account as represented in this post seems unable to answer some salient questions on the nature of that mechanism. A more detailed analysis of the effect of different accounts of causation on the HPC theory as well as an evaluation of Boyd’s semantics would seem to be the next step, but that is a task left for another post.

References Beck, A. T., Alford, B. A. 2008. Depression: Causes and Treatment. University of Pensilvania Press. Borsboom, D. 2008. Psychometric Perspectives on Diagnostic Systems. Journal of Clinical Psychology 64: 1089-1108. Boyd, R. 1991. Realism, Anti-Foundationism and the Enthusiasm for Natural Kinds. Philosophical Studies 61: 127-148. Boyd, R. 1999. Homeostasis, Species and Higher Taxa. in Wilson, R. A. ed. Species: New Interdisciplinary Essays. MIT Press. Boyd, R. 2010. Realism, Natural Kinds and Philosophical Methods. In Beebee, H., Sabbarton-Leary, N. eds. The Semantics and Metaphysics of Natural Kinds. Routledge. Chakravartty, A. 2007. A Metaphysics for Scientific Realism. Cambridge. Craver, C. F. 2007. Explaining the Brain. Clarendon Press. Devitt, M. 2008. Resurrecting Biological Essentialism. Philosophy of Science 75: 344-382. Ellis, B. 2001. Scientific Essentialism. Cambridge. Kendler, K. S., Zachar, P., Craver, C. 2011. What Kinds of Things are Psychiatric Disorders? Psychological Medicine 41: 1143-1150. Kendler, K. S. 2008. Explanatory Models for Psychiatric Illness. American Journal for Psychiatry 165: 695-702. Kornblith, H. 1993. Inductive Inference and its Natural Ground. MIT Press. Putnam, H. 1975. The Meaning of Meaning. In Putnam, H. Mind, Language and Reality: Philosophical Papers. Cambridge. Reydon, T. 2009. How to fix Kind Membership: A Problem for the HPC Theory and a Solution. Philosophy of Science 76: (5) 1-19. Slater, M. H., Borghini, A. 2011. Introduction. in Campbell, J. K., O’Rourke, M., Slater, M. H. eds. Carving Nature at its Joints. MIT Press. Strevens, M. 2008. Depth. Harvard University Press. Wilson, R. A., Barker, M. J., Brigandt, I. 2007. When Traditional Essentialism Fails: Biological Natural Kinds. Philosophical Topics 35: 189-215. Wilkerson, T. E. 1995. Natural Kinds. Ashgate Publishing. Woodward, J. 2002. What is a Mechanism? A Counterfactual Account. Philosophy of Science 69: 366-377.