An ideal disorder? Autism as a psychiatric kind

Abstract

In recent decades, attempts to explain autism have been frustrated by the heterogeneous nature of its behavioral symptoms and the underlying genetic, neural, and cognitive mechanisms that produce them. This has led some to propose eliminating the category altogether. The eliminativist inference relies on a conception of psychiatric categories as kinds defined by their underlying mechanistic structure. I review the evidence for eliminativism and propose an alternative model of the family of autisms. On this account, autism is a network category defined by a set of idealized exemplars linked by multiple levels of theoretically significant properties. I argue that this network model better captures the empirical phenomena, the historical growth of the category, and the ways the category has been shaped by social norms and interests. Finally, I defend a realist interpretation of network categories against the challenge from eliminativists.

Keywords:

• autism
• mental disorder
• psychiatry
• taxonomy
• classification
• nosology

Acknowledgments

Thanks to Kathryn Tabb for sharing her forthcoming work and for helpful discussions of this material.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes on contributor

Daniel Weiskopf is an Associate Professor of Philosophy and Associate Faculty of the Neuroscience Institute at Georgia State University. He is the author, with Fred Adams, of An Introduction to the Philosophy of Psychology (CUP 2015). He has written numerous articles on concepts, embodied and extended cognition, and modeling in cognitive science, and he is presently working on a book on concepts and higher cognition tentatively entitled The Vehicles of Thought.

Notes

1. Although I mostly refer to autism as a “disorder” throughout this discussion, I recognize that there are those who reject this label, preferring the language of difference instead (Silberman 2015). In some cases, at least, it is clear that persons with autism can display patterns of abilities that are significantly superior to those of neurotypical individuals, and it is a complicated question whether to regard these departures from the “norm” as disorders (Lorenz and Heinitz 2014).

2. Such at least is my impression; see Wilson, Barker, and Brigandt (2007) for evidence to this effect, and also Magnus (2012), Ch. 6. Note that the claim here is that philosophers tend to reconstruct theories and practices in the biological sciences in terms of the HPC model, not that scientific practitioners themselves tacitly endorse this model. Some have made this latter claim, though I have my doubts about it and for present purposes I leave it an open question.

3. It's widely acknowledged that the HPC account is less well-suited to many other sciences. Magnus (2014) notes, for instance, that it fails to capture the relations among the fundamental inventory of particles posited by the Standard Model in microphysics. Critics have also questioned whether it can account for non-causal, functional, and highly polymorphic kinds, many of which occur within biology (Ereshefsky and Reydon 2014), and even its applicability to species is not as clear as its proponents have advertised (Reydon 2006). For an overview of kinds in psychiatry, see Zachar (2014).

4. The term “mindblindness” has been used to cover a range of different capacities – see Gallagher (2004) for discussion.

5. The same sorts of considerations can plausibly be deployed against an even more modest thesis, namely Muhammad Ali Khalidi's causal theory of kinds, on which all that is necessary for kindhood is something's being a node in a causal graph (Khalidi 2013). Autism does not correspond to any single such node, as the model developed in the next section illustrates.

6. Perhaps not all psychiatric disorders work like this. Some might be much more straightforward. The present claim is only that heterogeneous network-based disorders, of which autism is a paradigm, constitute an important and interesting class that have been largely neglected. See Section 6 for further caveats and limits on the view.

7. While the category structure described here exists as a sociological phenomenon rather than as something represented within the minds of individuals, a number of studies show that people can readily learn and use categories that involve this sort of exemplar-driven growth, and that languages sometimes encode them (Heit 1992; Xu, Regier, and Malt 2016).

8. Cushing (2012) has charted several of these revisions to the diagnostic criteria for autism in recent decades, and argues not only that these shifts have not been motivated by advances in scientific understanding, but also that intrusions of value into the classification undermine any hopes of interpreting it realistically. As the next section makes clear, I disagree with this pessimistic assessment. Verhoeff (2014) offers a more detailed look at the history of attempts to stabilize autism across disciplines, reaching a similar but more tentative eliminativist conclusion.

9. This marks one of several ways in which the network model proposed here differs from that of Cramer et al. (2010), who straightforwardly take disorders to be networks of comorbid symptoms.