2.1 Narrowing the Field

Debate about addiction is often organized around the following question: Is drug use in addiction a choice or is it compulsive? We can thus refer to proponents of the compulsion and choice (or moral) models of addiction. As we will see, it is not difficult to reject an uncompromising version of a compulsion model. Relatively uncontroversial and established science, which I review in this section, suffices. It is also important to reject what we might call an extreme pharmacological view according to which exposure, perhaps even brief exposure, to addictive drugs alone is sufficient to cause addiction. This view will also be rejected presently. At the same time, the assumption that neuroscience I review in the following section is relevant for understanding addiction is inconsistent with the opposite of an extreme pharmacological view. We need a view which acknowledges that there is something about the pharmacology and neurobiology of drug use that is distinctive and powerful, but which does not reduce to a cartoonish view of addiction as literal compulsion. The role of choice, which is native to the domain of moral psychology, I will have much to say about in section 3.4 once the other aspects of the picture have been outlined.

Consider the extreme pharmacological view. The influential pharmacologist Avram Goldstein seems to endorse such a view in what follows:

If we arrange matters so that when an animal presses a lever, it gets a shot of heroin into a vein, that animal will press the lever repeatedly, to the exclusion of other activities (food, sex, etc.); it will become a heroin addict. A rat addicted to heroin is not rebelling against society, is not a victim of socioeconomic circumstances, is not a product of a dysfunctional family, and is not a criminal.³

Though this is a particularly strident expression of such a view, what Goldstein is saying is familiar to many who were exposed to public-facing anti-drug use policy statements from the ‘80s and ‘90s containing phrases such “[this drug is] so addictive once is enough”, and so on.

We can safely reject this view. Most people who use addictive drugs do not become addicted. Approximately 40% of Americans admit to having used an illegal drug in their lifetimes, but even the highest estimates put the addiction rate understood to include addiction to prescription pain medications and alcohol- somewhere between 8% and 10%. Further, most people who do abuse drugs at some point in their lives manage to stop. The U.S. National Institute of Mental Health (NIMH) conducted a landmark study between 1980 and 1985 to measure the prevalence of psychiatric disorders amongst the U.S. population according to the then-current DSM III criteria, the Epidemiologic Catchment Area Study (US DHS 1994). One of the study’s most striking findings is that more than half of those who previously met the criteria for drug abuse or dependence reported no symptoms at all by age 24; by age 37, almost 75% are symptom-free.

Another consideration which appears to be inconsistent with the extreme pharmacological view comes from questioning the experimental paradigm used in earlier animal studies on addiction. Most notable among researchers pursuing this line is Bruce Alexander:

We compared the drug intake of rats housed in a reasonably normal environment 24 hours a day with rats kept in isolation in the solitary confinement cages [...]. This required building a great big plywood box on the floor of our laboratory, filling it with things that rats like [...]. The rats loved it and we loved it too, so we called it “Rat Park”. ^{(2010, p. 3)}

Alexander and his colleagues found that compared with animals housed in Skinner boxes, the rats in Rat Park took morphine at dramatically lower levels. They also found that this held for rats bred to have a metabolic dependence on morphine at birth (^{Alexander et al. 1981}). This suggests that the addiction of the caged rats to morphine is not caused by exposure to morphine, which they have in common with the rats in Rat Park, but by something else.⁴

We can also reject the extreme compulsion view. That view says that persons living with addiction possess very little or no control over their drug-taking. Though such a view is indeed extreme, it is by no means fringe. For instance, Nora Volkow, current head of the National Institute on Drug Abuse (NIDA), a $1 billion U.S. federal agency, appears to take such a view, calling addiction a “fundamental” disruption to “self-control” ⁽²⁰¹⁵⁾. Moreover, when researchers and clinicians characterize addiction as a disease (a controversial but by no means fringe position; see section 4 for further discussion), they often intend to invoke compulsion as the hallmark of such a disease. As Heyman and Mims report, “when addiction specialists say that addiction is a disease, they mean that drug use has become involuntary” ^{(2016, p. 386)}.

Notice that taken together, the extreme compulsion view and the extreme pharmacological view are especially implausible. They imply that mere exposure to drugs can cause a fundamental disruption to one’s capacity for self-control. The view in this form is refuted by ordinary experience.

The extreme compulsion view on its own would appear to be refuted by the fact that persons with addictions can modulate their drug-taking behavior in response to incentives. For instance, when physicians in treatment for addiction are randomly tested and threatened with job loss if they relapse, abstinence rates remain as high as 80-90% (^{Ganley et al. 2005}; ^{Bohigian et al. 2005}). Smaller incentives also show impressive effects. Further, if those who are incentivized to remain abstinent with vouchers that they can exchange for modest goods are compared to controls who receive only counseling, those in the voucher program consistently do better (^{Higgins et al. 1991}; ¹⁹⁹⁴; ¹⁹⁹⁵). The most natural interpretation of this is that subjects are exercising self-control in these settings.⁵

The epidemiological and environmental considerations adduced against the extreme pharmacological view also have force against the extreme compulsion view. If addiction involves a “fundamental” disruption to self-control, why do most people with addictions “mature out”? And how is it that rats who have a morphine “addiction” are able, given the right environment, to refrain from taking the drug when it is readily available? One possible response is that somehow all of these creatures just fail to have true addictions. But this threatens to cause us to lose an independent handle on the phenomenon. It requires thinking that addiction must be more than, e.g., morphine dependence from birth, or that it is not reliably tracked by the behavioral criteria used in the Epidemiologic Catchment Area Study. That may be true, but until the reply is buttressed with an independent positive characterization of addiction such that those who mature out are not true addicts, it seems to involve helping itself to what is under dispute.⁶

2.2. Neuroscience

Let us now consider the neuroscience of addiction. The goal in this section is to draw valuable lessons that can be applied to the moral psychology of addiction. Addictive drugs all increase dopaminergic activity⁷ and three of the brain’s main dopaminergic pathways, the mesolimbic and mesocortical pathways, and the nigrostriatal pathway, are widely thought to be implicated in addiction. Crucially, it is also well-established that the mesocorticolimbic system not only responds to rewards, but to cues that reliably predict rewards. (^{Schultz et al. 1992}; ^{Shultz et al. 1997}). In recent years, two major theories have been advanced regarding the role of dopamine neurotransmission in addiction: (1) the prediction error theory and (2) the incentive sensitization theory. These theories are often presented as competing (^{Hu 2016, pp. 300-301}; ^{Berridge 2007}).⁸ Against this, I will argue that they should be taken to be complementary.

3.1. Addiction as a “Defect of the Will”

Let us begin by taking Wallace’s view as representative. His view is meant to capture the sense in which addiction is what he calls a “defect of the will”. Wallace argues that motivational states come in two varieties, viz., those that are merely given to us, and those that are the result of primitive episodes of agency such as choice, decision, or the formation of an intention. The latter show deliberative and volitional capacities at work, whereas the former come unbidden and can be obstacles to acting in accordance with considered judgment. When we add to this picture the ordinary notion of belief, we get a tripartite moral psychology.¹² As we will see, in addiction all three elements are present but distorted.

In Wallace’s formulation, what makes drug addiction practically challenging is that it involves particularly strong and persistent motivational states of the merely given variety. Wallace says:

On the [...] account I have offered, [desires to take drugs] involve the intense focusing of one’s attention onto the anticipated pleasures of (say) drug consumption [...]. But someone subject to such a quasi-perceptual state will presumably find it difficult to think clearly about the overall balance of reasons bearing on the decision to consume or abstain from consuming the drug. Let us suppose [the balance of reasons favors abstaining.] Adding such a desire to the mix [...] would make it much harder for the agent to reach this conclusion and to keep it firmly in view. The focusing of one’s attention onto the pleasures of consumption [distorts judgment], and this distorting effect can be considered an impairment of the agent’s capacities for practical rationality. (^{1999a, pp. 645-646})

The neuroscience suggests that Wallace is correct to emphasize the way in which addictive motivations are persistent and operate primarily by grabbing attention.¹³ But the same neuroscience suggests that he is overemphasizing, or perhaps mislocating, the role of pleasure. Indeed, Wallace explicitly links pleasure and salience by adopting a “phenomenological” conception of desire, according to which at least the desires operative in addiction represent, in a quasi-perceptual mode, the action of drug-taking as pleasant (1999a, pp. 641-643).

But salience can be independent of pleasure or anticipated pleasure. People with addictions often report wanting to use drugs even though they know it won’t be pleasant. Insofar as the operative motivational state in addiction is a cued intrinsic desire for drug taking of the variety posited by the incentive sensitization account, this is not surprising. On that model, not only needn’t the action of drug-taking be represented as pleasant, its satisfaction also needn’t ultimately be pleasant, as the distinction between liking and wanting suggests.

Some theorists also think that there may be a distinctive role for hedonic dysregulation in some phases of addiction and not others. As Meyer et al. put it: “[H]edonic allostasis may maintain drug-taking behaviors during early withdrawal or attempts to reduce drug use, whereas incentive sensitization may promote relapse even after physiological and emotional withdrawal symptoms have subsided” ^{(2016, p. 478)}. This could mean that pleasure and relief from distress are involved in relapse in early abstinence but not in later abstinence.

In any case, the role of pleasure and relief from distress assuming that these are two sides of the same conceptual coin- is not as straightforward as Wallace supposes. Where desire to take drugs does not represent that course of action as pleasant we need to know: What is the mode of presentation characteristic of such a desire?

3.2. Hybrid Intentions

An additional necessary refinement to Wallace’s view points the way. Wallace is right to contrast desires which are simply given to us with those motivational states which are up to us, and he naturally thinks that, on this dimension of comparison, addictive desires should be grouped with the former. However, a leading and natural interpretation of the incentive sensitization theory has it that addictive desires are like the latter in that they lead directly to action, even if they are unlike them in not being up to us:

[A]ddictive desire does not typically function like [an ordinary desire]. It does not serve as an input to deliberation, something to be weighed, along with other competing desires, in deciding what to do. Instead, addictive desire functions as something more like an intention: as something that, unless checked, will lead, in a rather direct way, to action. (^{Holton and Berridge 2011, p. 241})

Wallace understands the contrast between the two kinds of motivation to be a contrast between what is merely “hydraulic” forces to which we are largely passive bystanders- and what are intimate products of our wills. Paradigmatically, intentions are like the latter, but if addictive motivations operate as suggested here, the operative states in addiction might not be desires at all, but instead deviant intentions which merely assail us.

One could attempt to make a functional distinction between the intention-like states we find in addiction and true intentions to preserve the tidiness of Wallace’s distinction. Perhaps, for instance, it is a functional requirement on intentions that they be linked in the right way to the exercise of deliberative and evaluative capacities.

This move is attractive, but it rests on a refusal to acknowledge deviant members of a functional type which would be very costly in the present context. Consider what a traditional functionalist might have to say about, e.g., a defective heart. If we consider only what the heart in fact does to determine what functional type it belongs to, we might be forced to say it is not a heart at all, but this doesn’t seem like the right thing to say. Teleofunctionalists in the philosophy of biology (^{Neander 1991}; ¹⁹⁹⁵) capture this by saying that a defective heart counts as a heart even though it cannot pump blood because it is an entity of a type whose function is to pump blood. This fact about function, in turn, is determined by evolutionary history. It was to pump blood for which entities of the heart type were selected. This is as true of the defective heart as it is of the efficient heart.

Addiction seems to involve dysfunction at multiple levels. If we adopt teleofunctionalism, these dysfunctions need not be type-disqualifying. This is attractive if we want to avoid a proliferation of new functional types up and down our analysis of addition. Moreover, it offers an attractive way of thinking about the deviant intentions which the incentive sensitization theory posits. If intentions are generated by certain cognitive mechanisms, and those mechanisms have the right kind of evolutionary history, the product of those mechanisms will count as intentions, on teleofunctionalist grounds, even if they don’t have the complete functional profile of typical intentions.

If this is correct, then the intentions characteristic of addiction straddle the standard boundaries of tripartite moral psychology. I shall therefore call them hybrid intentions. They are intimately related to agency on the front end, so to speak, even if the connection with volition that is typical of intention on the back end is not present. The result is a state which drives action in accordance with one’s internal states, but which is not responsive to the will.¹⁴

To further flesh out the difference between paradigmatic intentions and hybrid intentions, consider some of the features that are standardly attributed to intentions. Intentions preserve the motivational force of a choice or decision for later, resisting, at least to a fairly high degree, revision. ^{Bratman (1987, p. 16)} calls this feature of intentions stability. Typical intentions are also what ^{Bratman (1987, p. 16)} calls controlling. Holton, echoing Bratman, says, “The agent forms the intention at one time either by making an explicit conscious decision to perform the action or by some less deliberate, more automatic, process. Then, unless it is revised, the intention will lead the agent to perform the action directly” (^{2009, p. 2}; emphasis in original).

As Holton says, it is not necessary that intentions be arrived at through conscious deliberation. The more automatic processes that can lead to intentions might include unconscious deliberation (if one believes in such a thing) or might involve the output of certain cognitive modules which are relevant for action and evaluation. It is important, however, that they are responsive to reasons. Perhaps practical judgment can be made unconsciously, or can partly arise from subpersonal processes. But if one were to revise one’s judgment, one’s intentions would (ideally) change accordingly; and without revising such a judgment, intentions, it seems, cannot change. One could say, borrowing a term from ^{Scanlon (1998)}, that intentions are “judgment-sensitive” attitudes.

So, if there were to be a state that didn’t reflect judgment¹⁵ but was nevertheless stable and controlling, it could be a liability with respect to acting well. It would be the sort of state whose formation is not subject to volitional control, as typical intentions are, but which possesses a high degree of motivational efficacy which persists over time. It would be a (defeasibly) action-controlling state over which its subject would have, in turn, limited control.

The sort of motivational states invoked by the incentive sensitization model appear to be states of just this kind. They come on the scene in direct response to a cue and so they are not the result of deliberation, conscious or unconscious. Moreover, they are also unlike typical intentions which arise from sub-personal mechanisms (if such there be) because they can’t be overthrown with a revision of judgment. All the same, the connection between them and action is very close. Unlike desires, they don’t serve as inputs to deliberation. Rather than partially contributing to how practical questions are answered, they represent such questions as having already been decided. Their motivational force is also persistent and, because the process by which drug-cue relationships are established works by mere association, almost anything that can be associated with a drug can operate as a hybrid- intention-triggering cue.

As with typical intentions, the presence of a hybrid intention does not, of course, guarantee that the corresponding action will be performed. However, it is much more difficult to overthrow a state which is not under volitional control than it is to overthrow one which is. If a state is not judgment-sensitive in the first place, it may be impervious to changes (or retrenchments) of judgment, no matter how clear or forceful.

If this is how hybrid intentions function, and hybrid intentions are central to motivation in addiction, it would be a mistake to think that trying to act well in the face of such motivations can be understood on the model of struggling against a desire. Watson seems to be getting at something similar when he says:

We will do well, I think, to abandon [a view of addiction according to which we center] the power of addictive desire to defeat our best efforts and, instead, to understand the relevant notion of compulsion in terms of the tendency of certain incentives to impair our capacity to make those efforts. We are not so much overpowered by brute force as seduced. (2004a, p. 71; my emphasis.)

^{Burdman (2024)} also emphasizes that we should not think of the addicted person as struggling “against a force she cannot oppose” (p. 58). But it is unclear that we can make the required shift to the recommended picture without recognizing the way in which intentions in addiction are hybrid states. Being seduced is, I take it, to be understood in terms of the fact that intentions in addiction constitute part of the agent’s practical standpoint, the point of view from which action springs, as typical intentions do. Yet, they do so without being subject to the will. That is, typically, volitional states like intentions and choices are closer, logically and causally, to action than desires. Hybrid intentions share this property. By contrast, desires of the merely given variety are not themselves deliberable and are not, except where they are instrumental, the outputs of deliberation. Hybrid intentions share this property. One could say that hybrid intentions (though it is admittedly anti-Anscombean to think of them as mental states) correspond to all three of Anscombe’s forms or guises of intention: they concern the future, they pick out that for the sake of which the action is done, and they make action done accordingly intentional action ⁽²⁰⁰⁰⁾. They are therefore action-theoretically thicker than mere desires, though because they are judgment-insensitive, they can lack a hallmark of typical intentions, viz., “an all-out, unconditional judgement that the action is desirable” (^{Davidson 1980, p. 99}).

I take this to suggest a somewhat more sophisticated picture of how addiction is a volitional impairment than the one with which we began. Merely being subject to states with this anomalous functional profile is a volitional vulnerability simply because one’s overall inventory of non-deliberable states is thereby expanded, and expanded by the addition of states which bear a direct connection with action. Perhaps more importantly, insofar as an agent’s action-facing states constitute her practical view on the world, that practical orientation is compromised when it is infiltrated by states which are not responsive to her judgments or valuations.

3.3. Unstable Representations and Cognitive Distortion

Wallace assumes that representational states in addiction are not dysfunctional. But the prediction error theory suggests that this is not so. It implies that people with addictions cannot accurately represent the value of drug-taking, i.e., that the very representation of the value of drug-taking is manipulated by the pharmacological action of drugs to be ever-increasing.

Since this means that some of the cognitive materials on which practical deliberation takes place are themselves distorted, we should expect this to have downstream effects in action and to have effects elsewhere in the agent’s practical decision network, manifesting in both a biasing towards the action of drug-taking and the spreading of cognitive distortion to states which are in an instrumental relation to that action.

We should also expect such distortion to be a springboard for various other cognitive biases. We know that biases such as the confirmation bias recruit cognition in defense of representational states. ^{Pickard (2016)} highlights role of denial in addiction, but she also makes clear that denial is only one manifestation of the face of addiction as a “disorder of cognition” (p. 278). Other aspects include: “information-processing biases, motivational influences on belief formation and self-deception, and cognitive deficits with respect to insight and self-awareness”¹⁶ All of these processes are directly related to the artificially positive representations which the prediction error theory predicts we should find that those with addictions have towards drug-taking.

Other views in the philosophical literature also model addiction at the level of belief or cognition. For instance, ^{Levy (2014)} understands addiction as involving oscillations between all-things-considered judgments for and against drug-taking caused by competition between top-down and bottom-up processing. ^{Sripada (2022)} understands addiction to involve unreliable control over automatic thoughts which contribute to a distorted understanding of the true value of drug-taking. Levy’s model makes explicit appeal to the prediction error theory in explaining why the oscillation occurs: minimizing prediction error in the presence of drug cues leads to a model in which drug-taking is highly valued; but minimizing prediction error under other circumstances leads to the judgment that abstention is more valuable. Sirpada’s does not, but insofar as his view posits an inability to form a stable judgment of the value of drug-taking, seems consonant with it.

One of the chief advantages of the approach that I am taking here and one of the key payoffs of arguing, however tentatively, for the complementarity of the prediction error theory and the incentive sensitization theory- is that I can recognize the role played by prediction-error-caused cognitive distortions, while also assigning a central role to what seems to be an ineliminable and distinct aspect of addiction, viz., volitional and motivational dysfunction. If this is correct, competing views that do not recognize the role of hybrid intentions are leaving out a key piece of the picture.

3.4. Choice, Agency, and Responsibility in Addiction

Acknowledging all of these ways in which addiction affects the ordinary economy of judgment, motivation, and volition is consistent with the claim that those who have addictions choose to take drugs. But this is a rather superficial truth. The impairment of addiction does not show itself in how it bypasses choice, it shows itself in how the capacity for choice is re-directed. If choice and other deliverances of the will are the bottlenecks through which practical agency ultimately reaches action generally, this is what we should expect.

Nevertheless, we should reject certain choice-based models of addiction, such as ^{Heyman’s (2009)}. Heyman’s view is sophisticated, but in sum he thinks that addiction can be modeled using the tools of rational choice theory and behavioral economics. According to Heyman, from a local-choice perspective, drug-taking will often be the rationally superior choice even if, from a global-choice perspective, abstinence or reduction is clearly warranted. Addiction, then, involves making locally rational but globally irrational choices.

It is true that there are often clear temporal and probabilistic disparities between the value of drug-taking when assessed from the global- and local-choice perspectives. As Heyman says,

From the perspective of current choice, specious rewards [like drugs] have high value (because of immediate benefits and hidden costs), whereas from the perspective of global choice, their effective value is their true value because the costs have as much weight as the benefits. (2009, p. 146)

Some natural rewards (such as fatty and sugary foods) are specious in this sense and are familiar as such. But drugs have several properties which natural rewards lack which manipulate the local choice architecture and bias towards local-choice perspective-taking. For example, drugs undermine the value of competing choices by being “behaviorally toxic” (^{Heyman 2009, p. 145}). Typically, when we attend to a particular activity, the value of doing other things will increase, perhaps by becoming more pressing, or as alternatives become more interesting. Drugs often don’t work like that. Instead, they continue to grab attention and the effects of withdrawal and intoxication make other things less rewarding. This tilts the comparative judgment between locally available options in favor of drug-taking.

Heyman’s analysis illuminates the ways in which drugs manipulate choice architecture and set up a framework within which drug-taking choices are locally rational. However, the framing of addiction fundamentally through the lens of choice can start to look misguided once the neuroscience is under our belts. We needn’t dispute that there are intentional states we can attribute to agents in virtue of which their behavior would appear to be locally rational. But if the ultimate source of such states is direct manipulation by the pharmacological action of drugs and/or resulting neuroplasticity, it is unclear what the significance of the availability of this interpretation is supposed to be.

If pharmacology and neurobiology are at the root of the observed behavior, asking whether such behavior can be given a rational interpretation is not altogether probative. Doing so simply involves taking for granted the valuational and motivational states which only arise because of drug-induced neuroplasticity and asking what is rational relative to them. But such a perspective overlooks the way in which addiction involves disruptions to the very states and processes against which assessments of local rationality take place.¹⁷

In contexts where an agent is in full command of her cognitive and volitional capacities and is fully informed about the possible costs of acting in a certain way, choices bear an intimate connection with moral responsibility. Paradigmatic choices are made on the basis of the agent’s assessment of the balance of reasons. As such, inability to successfully justify a choice can occasion a number of different responses and forms of negative moral appraisal.

But the connection between choice and the appropriateness of these responses is unclear in addiction. Choices to take drugs can either fail to reflect the agent’s assessment of the balance of reasons (if action follows directly from a hybrid intention) or fail to reflect a competent assessment of those reasons (if based on distorted cognition). Choices paradigmatically render agents accountable,¹⁸ in the sense of being the appropriate targets of reactive attitudes and other sanctions, but this is unclear when the choices have the properties that they have in addiction.

Nevertheless, we should notice that two kinds of assessment still seem to have relatively unproblematic purchase. First, the states that impair cognition and volition in addiction do not operate in isolation from other things the agent believes, wants, or is committed to. The fact that addiction does not involve literal compulsion is enough to show this. Those with addictions respond positively to therapeutic support, to well-crafted choice architecture, to the cultivation of self-insight, and perhaps more crucially, to being held accountable to a community (^{Pickard and Pearce 2013}). All of this only seems appropriate if we see addiction broadly within the framework of agency. The fact that recovery is possible under such conditions suggests that failure to do so soon enough or effectively enough could render an agent negatively assessable¹⁹ if enough supportive background conditions are satisfied.

Second, the action-facing states which are triggered by drug cues and the (distorted) assessment of the value of drug-taking are still states of the agent. Thus, if instead of asking whether it would be fair to blame an addict for her conduct, we engage in assessment of her character roughly, one’s more-or-less stable set of cognitive, behavioral, and affective dispositions- we may find an appropriate target for moral assessment. The states in question are states for which the agent is not fully responsible, but in general that is not a condition on being an object of aretaic assessment (^{Watson 2004b}), as evidenced by the banal fact that no one is wholly self-created (^{Wolf 1987}) and that we nevertheless (rightly, it seems) take the ways people are to have great interpersonal significance (^{Smith 2005}).

So, while the psychological states and processes characteristic of addiction may not be such that the agent is fully accountable for them and their behavioral upshots, they may nevertheless be attributable to the agent (^{Watson 2004b}). They are states of the agent and because of their stability and (in severe cases) relative centrality to the character of the agent, ground certain judgments about that character.

This is relevant not only because it allows a certain kind of assessment to get a foothold. It is also only because those states are attributable to the agent that they can come under her indirect control and can be tempered, modified, inhibited, or displaced given the other traits of character and states of mind that constitute her practical self. Indeed, the degree to which the subject ultimately succeeds in taking control of her life depends on the rest of her character being brought to bear in this way. It is thus no surprise that treating those who are addicted as responsible that is, as agents who face obstacles to their agency but who nevertheless have the power to reform- is a promising therapeutic avenue for recovery (^{Pickard and Pearce 2013}).

Although there are a great many complications about addiction and moral responsibility that must remain off the page, I take myself to have pointed the way to the following:

I take these conclusions to be significant, in part, because of the way they relate agency in addiction to non-disordered agency. Agency in addiction isn’t just agency in the face of unruly appetites. But it is also not merely being pushed around mechanically by states which bear no relation to the states recognized by ordinary psychological explanation. With the major signposts on the terrain, we can begin in earnest to tease out the ultimate significance of addiction in context.