First, the praise. Redish et al. show us the inherent variability in decision-making. That is a fundamental Darwinian insight, at a remove from the singular focus of the traditional reward paradigm in psychology or prominent evolutionary approaches such as optimality or adaptive modules (Lende Reference Lende, Trevathan, Smith and McKenna2007; Lende & Smith Reference Lende and Smith2002). Redish et al. accomplish this feat by tying together neurobiological, psychological, and animal research – meaning we no longer face the “black box” of the brain or an over-reaching theory based on one slice of research. Redish et al. build their analysis of addiction on similar strengths – an emphasis on variability in vulnerabilities, comprehensive appraisal rather than a pet theory, and the much-needed linkage of decision-making to different addictive drugs and activities.

Unfortunately, three fundamental problems undermine their overall efforts. First, Redish et al. base their approach on the “rational man” assumption. As they note, the literatures they use “have converged on the concept that decisions are based on the prediction of value or expected utility of the decision” (sect. 3, para. 1; emphasis theirs). Expected utility lies at a considerable distance from addiction, where excessive involvement in a behavior, social role failure, and the active ignoring of costs are core diagnostic features.

Second, Redish et al. approach decision making as largely innate. Their computational framework works through internal mechanisms, a series of calculations based on expected utility, with the twist thrown in of a fast habit system and a slower planning system. However, both brain research and decision-making research have shown that interactive, environmentally dependent processing are fundamental features of adaptive behavior (Chiel & Bier Reference Chiel and Beer1997; Clark Reference Clark1997; Epstein Reference Epstein2007). This embodied approach stands in direct contrast with a representational (or computational) view of brain function, and it undercuts Redish et al.'s claim that they have presented a unified framework.

Third, Redish et al. do not draw on either ethnographic or clinical case studies of addiction, which hampers their ability to find creative connections between decision-making research and addictive behavior. This point is especially clear late in the target article, when they write, “the decision-making theories discussed in this article are primarily about reinforcement (delivery of unexpected reward) and disappointment (non-delivery of expected reward)” (sect. 6, para. 3). Unexpected rewards and non-delivery do not add up to addiction.

For example, with gambling, Redish et al. emphasize Vulnerability 6 (the misclassification of situations and the illusion of control). Using situations (rather than stimuli), neural plasticity, and the cognitive dynamics of “control,” the authors bring a novel interpretation for why addicts get in such a rut about recognizing the costs of their behavior. However, this interpretation builds from their ability to get away from a rational innatist position, precisely through situations, plasticity, and illusion.

Thus, implicit assumptions of rationality and innate computation hamper the development of a unified approach to how decision-making plays into addiction. Addiction involves subjective, sociocultural, and evolutionary elements, and these elements help us get at the “why” of addiction (Lende Reference Lende2005a; Reference Lende, Trevathan, Smith and McKenna2007). Rather than assuming that rationality explains why, more fruitful approaches can be achieved by combining anthropology and other social sciences with the biological and behavioral sciences (Hruschka et al. Reference Hruschka, Lende and Worthman2005).

Why do people use and abuse drugs? Environmental inequality makes a direct impact on decision-making (Bourgois Reference Bourgois2002; Singer et al. Reference Singer, Valentín, Baer and Jia1992), for example, in social position, dopamine function, and cocaine use (Morgan et al. Reference Morgan, Grant, Gage, Mach, Kaplan, Prioleau, Nader, Buchheimer, Ehrenkaufer and Nader2002), or in the quality of the environment and the reinforcement of cocaine and heroin (Alexander Reference Alexander1987; Alexander et al. Reference Alexander, Beyerstein, Hadaway and Coambs1981). Similarly, moving beyond the individual misclassification of situations, sociocultural context can directly shape heavy drinking, particularly the effect alcohol has and the amount that people drink (Heath Reference Heath2000; MacAndrew & Edgerton Reference MacAndrew and Edgerton1969; Marlatt Reference Marlatt, Peele and Grant1999).

Moreover, the understanding of relapse is enhanced through incorporating elements of self-efficacy, cognitive dissonance, and personal attribution (Brownell et al. Reference Brownell, Marlatt, Lichtenstein and Wilson1986). Rather than blaming the situation, the person blames “personal weakness,” heightening the chance of further relapse due to continued “loss of control” (after all, already lost control once …). In other words, relapse is more than a decision-making vulnerability based on the computation of benefits or costs in any particular situation.

Finally, drug users seek things from substance use – intentions, subjective experience, and meaning all matter in what users want. For example, heavy methamphetamine users often engage in “functional use,” seeking to enhance a skill or to be in an altered state while still engaging in socially acceptable behavior (Lende et al. Reference Lende, Leonard, Sterk and Elifson2007). Craving, long seen as psychobiological, is often defined by users in reference to personal control, rather than being separable into a cue-, drug-, and withdrawal-driven typology (Bruehl et al. Reference Bruehl, Lende, Schwartz, Sterk and Elifson2006).

Instead of reducing the habitual and compulsive aspects of addiction to computational decision-making, these behaviors and experiences can drive addiction. A renewed focus on the psychology of interest (Silvia Reference Silvia2008) and the use of wheel running as a behavioral model for compulsive involvement (Rhodes et al. Reference Rhodes, Garland and Gammie2003; Reference Rhodes, Gammie and Garland2005; Sherman Reference Sherwin1998) show us an important insight often lost in experimental models: neurobiology and decision-making serve behavior, not the other way around.

A core part of any behavioral involvement, including addiction, is subjective experience. To take one example, Robinson and Berridge's (Reference Robinson and Berridge2001) incentive salience (discussed in Redish et al.'s Vulnerability 4, overvaluation in the planning system) depends as much on the meaning of drug use as it does on the action of the mesolimbic dopamine system (Lende Reference Lende2005b). Heightened salience is a measurable trait, using a psychometrically valid scale (Lende Reference Lende2005b). At the same time, both cultural symbols and an individual's sense of self impact what users desire and seek out. In my ethnographic research, one girl who smoked marijuana nearly every day explained what she wanted: “estar en un video” (“to be in a video”), where attention was shifted away from how she felt in her traumatic yet culturally valued family environment. This type of compulsive involvement, both in itself and as a step away from a traumatic everyday life just waiting to return, is central to why some people use drugs to such excess.