The weight of theory and clinical evidence on mental disorders strongly supports the general perspective defended by Borsboom et al. in the target article. Indeed, increasing numbers of philosophers of science regard it as probable that most of the causal structure of the world, as identified across the sciences including fundamental physics, is best modeled in terms of networks. This is an important part of standard explanations for the fact that most scientific progress does not involve reduction (Horst Reference Horst2007). Scientific progress generally proliferates types of processes rather than collapses them (Ladyman & Ross Reference Ladyman and Ross2007), and the greatest challenge for most modelers is not isolating interrelated systems, but carefully entangling them in a way that still allows for rigorous estimation of parameters and causal effects in real empirical data.
The task that Borsboom et al. take on is explaining how and why network modeling undermines reduction in the specific case of psychiatry. Mental disorders, they argue, should seldom, if ever, be identified with brain disorders, though they recognize that the latter are often aspects of the former. However, they slightly undercook the radicalism of their challenge to reductionism as a clinical strategy. This is because their examples are limited to cases where the expression and perseverance of a complex brain state under a certain kind of mental characterization require reference to intentional patterns, which by their nature involve reference to environmental conditions and behavioral responses. But these are among the factors that block reduction of minds to brains generally; they fail to isolate particular obstacles to reduction that arise for disorders.
In each case that Borsboom et al. consider, one or more brain disorders are at least involved, even if not the “essence” of what makes the phenomena kinds of mental pathologies. To illustrate the full extent to which reductionism is misguided as a specifically psychiatric methodology, it is helpful to focus on a case in which the brain's contribution to a mental disorder results from its working just as natural selection “designed” it to do, but in an environment where this produces mental pathology. Gambling addiction is just such an instance.
Let us set the stage as conservatively as possible. Suppose that we apply the concept of “addiction” stringently, so that it refers, where gamblers are concerned, only to people who experience aversive somatic withdrawal symptoms when not gambling, to the extent that normal cognitive and emotional functioning in non-gambling settings is crowded out. This conceptual stringency might on the face of it seem to stack the odds in favour of a reductionistic account. It separates people who have a particular physiological dependency from the much larger population of problem gamblers who struggle with a range of social coping and self-management issues, and who lie along a smooth continuum with people for whom gambling is relatively low-cost recreation. The population of “true addicts” can be statistically distinguished from the wider group of “mere” problem gamblers (Kincaid et al. Reference Kincaid, Daniels, Dellis, Hofmeyr, Rousseau, Sharp and Ross2013), and such separation can be legitimately motivated by the suggestion that the former, but not the latter, are appropriately treated (in part) by neuropharmacological therapies (Ross et al. Reference Ross, Sharp, Vuchinich and Spurrett2008). Is this not an exemplary instance of a reductionist's response to a critique such as Borsboom et al.’s, wherein those suffering from a brain disorder are systematically carved out from a class of behaviorally related cases where social forces mimic the “real disease”?
As we add known causal details, the reductionist path seems at first to pay dividends. Addictive forms of gambling involve sequences of statistically independent events that sometimes yield rewards and can be generated by simple, stereotyped actions on the part of the gambler. The addictive syndrome is made possible by a conjunction of three facts, which are indeed about neural mechanisms and architecture:
1. Allocation of cognitive-emotional attention and reward prediction are approximately fused in the same neural learning circuit based in the ventral striatum.
2. The circuit in question operates a specific form of Rescorla-Wagner learning algorithm that cannot settle on a model of genuine randomness, preventing it from learning that there is nothing to be learned from repeatedly playing things like slot machines.
3. Once the wider system in which the striatal learning mechanism is embedded discovers an action pattern that reliably delivers information relevant to trying to predict the reward sequence, it automatically cues motor preparation for harvesting activity around the focus of attention. Somatic cravings are the subjective experience of such motor preparation in the absence of opportunity for action. The cravings cue thoughts about gambling, and the thoughts sustain attention and motor preparation; so there is self-sustaining feedback.
This story seems transparently to be mainly about the brain. But for medical purposes – that is, as an account of a disorder – it is not reductive at all. This is because every component of the causal network that lies within the cranium is working just as evolution selected it to do. An easily accessible learning opportunity is presented, and the appropriate neural network tries to take advantage of it in the way that has garnered sustaining reward streams for active vertebrates for hundreds of millions of years. Alas, cunning manipulators have engineered an environment in which this “proper functioning” (Millikan Reference Millikan1984) leads to emotional, cognitive, behavioral, and social disaster for people. Since the introduction of sophisticated digital technology into slot machine design, the proportion of casino revenues derived from addicts has exploded (Schüll Reference Schüll2012). There is no reason to hypothesize that any endogenous vulnerability in brains has recently increased; the entire cause of this psychiatric pandemic lies in malicious environmental engineering, plus regulatory failure. So we have an alarmingly widespread mental disorder, crucially reliant on neural properties of a causal network – and no brain disorder at all. There is not merely a failure of reduction but, contra a concession Borsboom et al. allow, even of supervenience, with respect to what is disordered – the mind (and society), but not the brain.
The weight of theory and clinical evidence on mental disorders strongly supports the general perspective defended by Borsboom et al. in the target article. Indeed, increasing numbers of philosophers of science regard it as probable that most of the causal structure of the world, as identified across the sciences including fundamental physics, is best modeled in terms of networks. This is an important part of standard explanations for the fact that most scientific progress does not involve reduction (Horst Reference Horst2007). Scientific progress generally proliferates types of processes rather than collapses them (Ladyman & Ross Reference Ladyman and Ross2007), and the greatest challenge for most modelers is not isolating interrelated systems, but carefully entangling them in a way that still allows for rigorous estimation of parameters and causal effects in real empirical data.
The task that Borsboom et al. take on is explaining how and why network modeling undermines reduction in the specific case of psychiatry. Mental disorders, they argue, should seldom, if ever, be identified with brain disorders, though they recognize that the latter are often aspects of the former. However, they slightly undercook the radicalism of their challenge to reductionism as a clinical strategy. This is because their examples are limited to cases where the expression and perseverance of a complex brain state under a certain kind of mental characterization require reference to intentional patterns, which by their nature involve reference to environmental conditions and behavioral responses. But these are among the factors that block reduction of minds to brains generally; they fail to isolate particular obstacles to reduction that arise for disorders.
In each case that Borsboom et al. consider, one or more brain disorders are at least involved, even if not the “essence” of what makes the phenomena kinds of mental pathologies. To illustrate the full extent to which reductionism is misguided as a specifically psychiatric methodology, it is helpful to focus on a case in which the brain's contribution to a mental disorder results from its working just as natural selection “designed” it to do, but in an environment where this produces mental pathology. Gambling addiction is just such an instance.
Let us set the stage as conservatively as possible. Suppose that we apply the concept of “addiction” stringently, so that it refers, where gamblers are concerned, only to people who experience aversive somatic withdrawal symptoms when not gambling, to the extent that normal cognitive and emotional functioning in non-gambling settings is crowded out. This conceptual stringency might on the face of it seem to stack the odds in favour of a reductionistic account. It separates people who have a particular physiological dependency from the much larger population of problem gamblers who struggle with a range of social coping and self-management issues, and who lie along a smooth continuum with people for whom gambling is relatively low-cost recreation. The population of “true addicts” can be statistically distinguished from the wider group of “mere” problem gamblers (Kincaid et al. Reference Kincaid, Daniels, Dellis, Hofmeyr, Rousseau, Sharp and Ross2013), and such separation can be legitimately motivated by the suggestion that the former, but not the latter, are appropriately treated (in part) by neuropharmacological therapies (Ross et al. Reference Ross, Sharp, Vuchinich and Spurrett2008). Is this not an exemplary instance of a reductionist's response to a critique such as Borsboom et al.’s, wherein those suffering from a brain disorder are systematically carved out from a class of behaviorally related cases where social forces mimic the “real disease”?
As we add known causal details, the reductionist path seems at first to pay dividends. Addictive forms of gambling involve sequences of statistically independent events that sometimes yield rewards and can be generated by simple, stereotyped actions on the part of the gambler. The addictive syndrome is made possible by a conjunction of three facts, which are indeed about neural mechanisms and architecture:
1. Allocation of cognitive-emotional attention and reward prediction are approximately fused in the same neural learning circuit based in the ventral striatum.
2. The circuit in question operates a specific form of Rescorla-Wagner learning algorithm that cannot settle on a model of genuine randomness, preventing it from learning that there is nothing to be learned from repeatedly playing things like slot machines.
3. Once the wider system in which the striatal learning mechanism is embedded discovers an action pattern that reliably delivers information relevant to trying to predict the reward sequence, it automatically cues motor preparation for harvesting activity around the focus of attention. Somatic cravings are the subjective experience of such motor preparation in the absence of opportunity for action. The cravings cue thoughts about gambling, and the thoughts sustain attention and motor preparation; so there is self-sustaining feedback.
This story seems transparently to be mainly about the brain. But for medical purposes – that is, as an account of a disorder – it is not reductive at all. This is because every component of the causal network that lies within the cranium is working just as evolution selected it to do. An easily accessible learning opportunity is presented, and the appropriate neural network tries to take advantage of it in the way that has garnered sustaining reward streams for active vertebrates for hundreds of millions of years. Alas, cunning manipulators have engineered an environment in which this “proper functioning” (Millikan Reference Millikan1984) leads to emotional, cognitive, behavioral, and social disaster for people. Since the introduction of sophisticated digital technology into slot machine design, the proportion of casino revenues derived from addicts has exploded (Schüll Reference Schüll2012). There is no reason to hypothesize that any endogenous vulnerability in brains has recently increased; the entire cause of this psychiatric pandemic lies in malicious environmental engineering, plus regulatory failure. So we have an alarmingly widespread mental disorder, crucially reliant on neural properties of a causal network – and no brain disorder at all. There is not merely a failure of reduction but, contra a concession Borsboom et al. allow, even of supervenience, with respect to what is disordered – the mind (and society), but not the brain.