- Commentary on “Epistemic Value Commitments”
The case for treating the underdetermination of psychiatric classification with just the same tools as are employed in solving the more general underdetermination of theory by data is well made by Sadler. Quite what that treatment amounts to, however, raises a number of issues that are not only central to any philosophical conception of the rationality of theory choice, but cut deep into issues about the appropriate model for rational clinical practice.
On the face of it, Sadler’s recommendation is plain and looks to clarify the central problems confronting the authors of DSM nosology. The central problem for the construction of DSM guidelines is that there is no algorithm for turning input data concerning observables (behavior patterns) into theoretical classification. A simple-minded positivism would have had us expect that the production of an adequate nosology by symptoms as a mechanical algorithm—plug in the data at one end and the nosology pops out the other end. The interesting philosophical issue concerns how we respond to this underdetermination. I want to note two stages to the response we make.
The first stage turns on deciding between thinking (a), that with sufficient further non-observational input some sort of modified algorithm can be produced for psychiatric classification, and thinking (b), that no such modified algorithm is available and that psychiatric classification irredeemably involves an element of judgment. The second stage that I want to explore concerns whether taking option (b) at the first stage seriously undermines the rationality of psychiatric classification and theory choice in general.
Sadler seems to favor option (a). He notes that some have responded to the Kuhnian diagnosis of the lack of effective decision procedure for the production of theory by impugning the possibility of any rationality to theory choice and by seeing the activity of theory choice as thoroughly soaked in value commitments that bear the hallmark of ideology rather than rationality. Sadler’s response, however, is to draw attention to a set of values that he calls “epistemic values” that are “intrinsic to the theorizing about and construction of the classification itself.” I see this as a version of what I have called option (a) because Sadler seems to treat these values as extra “bolt-ons” to the algorithm that turns observation into classification. His aim is to clarify the character of these non-observational constraints and then plot the way that their application renders tractable a current dispute in psychiatric nosology between categorical vs. dimensional classifications of personality disorders.
Suppose this is the option (a) strategy. The option (a) strategy involves a universalist model [End Page 227] of rationality. What I mean by “universalist” is this: A universalist model of rationality holds that if a transition from input to theory (input to classification) is rational, it is subsumable under a typology of such transitions that issues universal guides such as: Given inputs i 1, i 2 . . . under constraints c 1, c 2 . . ., conclude output/theory T n . A universalist model is distinct from a generalist model, for it accommodates the thought that the permissible routes from input to theory may include routes that are heavily conditioned by all sorts of particularities in the constraints cited in the transition rule. What the universalist does require, however, is that all rational transitions are instances of such a rule, even though the rule is not generalizable to cover all eventualities but is specific to certain particular circumstances. Sadler’s account of epistemic values is then an account that locates the place of these values as constraints that, in tandem with the inputs, determine output in a rule-governed manner.
There are a number of problems with this approach that I want to touch on before turning to the main philosophical moral that I would like to draw from Sadler’s paper. The universalist model is a model in which judgment plays no role in the transition from input to output. It is, however, unclear that such a model will really work on the evidence of Sadler’s investigations for a number of reasons. First, his initial definition of a “value” is that of a descriptor...