Redish et al. offer a unified framework of decision-making. They use the framework, first, to classify the different kinds of addictions; second, to show how the different theories of addiction are partial descriptions of these kinds; and, third, to prescribe different remedies for each kind. Addictions, they argue, arise from vulnerabilities in the access points of the system of decision-making – vulnerabilities similar to the errors and biases uncovered by behavioral decision research (Kahneman et al. Reference Kahneman, Slovic and Tversky1982).

Redish et al. propose a decision-making framework based on the distinction among three systems: “cognitive system,” which they leave out of the target article, and two action systems, namely, “planning system” and “habit system.” The planning system, or, a better term, “deliberation system,” involves assessing the values of outcomes (O) of a ray of actions (a), given the situation (S) (), and then choose the best (optimum) action. The habit system involves specific action in response to a situation, where outcome is absent (). The outcome is absent because the decision is very inelastic with regard to changes in the situation – which would demand different action if the decision is undertaken by the deliberation system.

It is a step in the right direction to ground addictions in decision-making theory. This commentary, though, finds that Redish et al. have failed to analytically distinguish between addictions, on one hand, and the biases and errors that inflict decision-making as highlighted by behavioral decision research, on the other. Further, this commentary questions whether Redish et al.'s framework is comprehensive enough to explain addictions. To start with, the habit system is not very different than the deliberation system: both are regulated by rational choice.

Redish et al.'s deliberation system resembles the standard notion of rationality. Agents are rational if they obey the transitivity axiom (consistency), completeness axiom (decisiveness), and some other minor axioms (see Kreps Reference Kreps1990, pp. 18–37). Further, agents must also undertake welfare-enhancing acts (see Becker Reference Becker1976, Ch. 1). But is the deliberation system, and its associated habit system, exhaustive description of behavior? Are organisms motivated only to take the best decision in light of a situation? Humans, for example, have the urge to act creatively and with imagination – that is, beyond what is suggested by the situation – which introduces innovations (Khalil Reference Khalil1997; Reference Khalil2007; Nooteboom Reference Nooteboom2000, Ch. 9). And such urge, or its frustration in the forms of depression and disorders, might be the basis for accounting for addictions. Although Redish et al. mention depression and disorders at the conclusion of the target article, they fail to incorporate them in their framework.

The main fabric of Redish et al.'s framework is rather the deliberation system, the habit system, and their interconnection. But habits do not lie far from deliberation. As the authors repeatedly state, habits originate from deliberation: When the organism faces repeatedly the same situation, there would be no need to deliberate; the organism would react automatically. This would speed decision-making, economizing on the use of cognitive resources. Such speeding is desirable as long as its benefit exceeds the cost of rigidity.

Despite the fact that habits originate from deliberation, they have different neural substrates. Redish et al., therefore, consider them dichotomous systems. “Because the association is a habitual, automatic association, choices driven by relationships will be unintentional, robotic, perhaps even unconscious” (sect. 3.3.1, para. 2). But is the habit system unintentional, robotic, or unconscious? The fact that they involve different neural substrates does not mean the two systems are dichotomous. They may involve an efficient division of labor. The same homogeneous structure differentiates itself into substructures, each specializing in a different function. So, the two systems might be underpinned by a common structure, rational choice.

In fact, the habit system does not lie far from rational choice. Let us use Redish et al.'s example of how driving to a new job becomes, after repetition, a habit (sect. 3.1). They recognize that such a habit never escapes the intervention of deliberation – as in the case when road construction closes a route. Nonetheless, aside from such interventions, they consider the habit system as autonomous. Let us assume that one's job moves to a nearby location, where halfway to work, the agent has to take another route. The agent would habitually fail to adjust midway, finding himself driving to the old job. But this repeated mistake does not go uncorrected. The old habit would eventually dissipate, given that the reward is suboptimal. Thus, rationality is, in the final analysis, a regulator of the habit system. We do not have a dichotomy. The habit system is a subsystem of deliberation.

If so, Redish et al.'s framework amounts to deliberation and its subsystem. They find that each system and its interaction with the others are full of vulnerabilities that explain a variety of addictions. They recognize that drugs enhance the vulnerabilities, leading to an over-evaluation of outcomes and probabilities. But they also argue that many of the vulnerabilities simply arise from errors in reasoning as uncovered by behavioral decision research (Kahneman et al. Reference Kahneman, Slovic and Tversky1982; Kahneman & Tversky Reference Kahneman and Tversky2000; Tversky & Kahneman Reference Tversky and Kahneman1981). Therefore, for Redish et al., addictions are analytically similar to errors and biases such as overconfidence, preference reversals, illusion of control, availability heuristic, conjunction fallacy, and so on.

However, there is a major difference between errors of judgment and addictions. Most agents commit the same errors of judgment in a predictable manner as they commit optical illusions (Ariely Reference Ariely2008). No such predictability exists, as Redish et al. admit, with regard to addictions. For instance, most agents are vulnerable to the switch from the loss frame to the gain frame in the Asian disease experiment (Tversky & Kahneman Reference Tversky and Kahneman1981). Also, most agents fall victim to overconfidence and the conjunction fallacy (Baron Reference Baron2008, Ch. 6). But, with addictions, individuals vary widely in the manner they may or may not become addicted.

The same decision framework seems unable to explain both biases and addictions. Redish et al.'s framework might not be the proper tool to explain addictions. Addictions, at first examination, are maladaptive actions in the sense that they reduce O. In contrast, the errors that arise from heuristics might be minor nuisances that the organism tolerates because the heuristics, on average, are efficient. In this case, the heuristics are tolerable “bad habits” given that such habits, in comparison to their absence, have positive net effect on O. Addictions, in contrast, totally diminish the ability to produce O. If so, we need another framework, aside from deliberation and habits, to tackle addictions. This framework may have to attend to the urge to be creative, to have a meaningful life, and how it may lead to addiction when the urge is frustrated.