Yesterday I posted about a new study that seems to shine some light on addiction to smoking, if “addiction” is defined based on the only clean definition I have seen in the literature, the Chicago School economic definition that focuses on increasing marginal(*) benefit from consumption. I noted that most of the time when “addiction” is used in scientific contexts, it as if it actually means something, but on closer inspection it does not. This facilitates the very anti-scientific behavior of using a word to mean one thing while letting the reader think it means something else.
[(*)As in “at the margin”=”at the edge”. It means the next bit, so in this context, it means the additional benefits from consuming the next bit. This should not be confused with other meanings of the word like “small” or “fringe”.]
An alternative way to use the term misleadingly is to define it to mean something that is idiosyncratic and absurdly different from what people normally interpret it to mean. This is typical in, for example, studies of youth smoking, where it is often defined as something like “indicates in a survey that he intends to smoke in the future”, which can mean that someone who has never smoked, or who smoked once and never did it again is “addicted”. (No, really, I am not kidding. The “research” by DiFranza et al. does exactly this, as described in this article and the commentary I wrote about it.)
Today, the New York Times decided to help me out with a great example of this method of misleadingly using the term. The asks “can exercise help curb addictions?”, which is pretty hard to answer if you do not know what the word means in this story. Fortunately, they define it. In this case, after providing the background — “according to an eye-opening new study of cocaine-addicted mice, dedicated exercise may in some cases make it even harder to break an addiction” — we learn that “break an addiction” means “stop bothering to come back to where the cocaine used to be offered once it is apparent that no more is coming”. It is explained that they are defining addiction in mice to mean “displayed a decided place preference for the spot within their chamber where they received cocaine”.
By that standard, I am addicted to the seat I usually take at the cafe, to say nothing of my desk chair. And the “addicted” mice might just be the ones who are slower to figure out that the world has changed, and so maybe it is common sense that is affected by exercise. In short, so far as addiction goes, this study only shows that your willingness to hang around and wait for cocaine seems to vary based on your exercise history, assuming you are a mouse. The only thing that seems particularly “eye-opening” was the effect of the cocaine on the mice.
The fatal flaw is that mice do not display sufficiently complex economic behavior that “addiction” makes sense for them. Thus, we end up with a lame metaphorical use of the term, in which it means nothing more than “demonstrates a desire to consume by waiting for the next delivery”. Trying to draw conclusions about addiction in mice is like trying to do experiments on mice to inform people about how to have a happier marriage. The conclusions from critter psych research often end up being right, but only because the truth about people is so clear that it creates a bias such that the researchers declare they are seeing evidence in the mice that supports the obvious facts about humans.
Actually, if you work at it, you can get some economic behavior out of mice and other critters. By forcing them to pull a lever a particular number of times to get a payoff, we can create a price, and thus with multiple levers and finite time, we have relative prices of various goods and a budget, the essential features of economic behavior. With that, adding in manipulation of the prices, it would theoretically be possible to measure whether the mice were displaying increasing marginal returns from consumption, a meaningful notion of addiction. But this would be quite difficult, and would require that the mice understood how the price was changing and other complications. It would also mean assuming that critters normally share the human trait of getting satiated with a good after consuming some and wanting to next consume something else instead, which runs contrary to my knowledge (i.e., arguably, they are addicted to everything they are positively disposed toward because they keep coming back to the same things).
So the mice researchers do not even bother trying to really measure economic behavior. However, it is interesting to note that their metaphorical use of “addiction” is still based on a behavior that is about as close to consumer economic behavior as is possible in the absence of prices. That is, they are implicitly recognizing my point from yesterday, that addiction is clearly a matter of economics rather than biology or epidemiology.
But let us imagine that a mice experiment with prices was created, and the mice displayed a willingness to “pay” (pull the lever) more for cocaine when they had used more recently, rather than getting less interested in it, whereas for goods that we assume are non-addictive, their willingness to pay followed the normal human pattern of dropping off as they consumed more. Would we still want to say they were truly addicted, as opposed to something like “acquired a very strong taste for it”. Something seems to be missing.
The colloquial notion of addiction has some element of the consumer not wanting to be in that state. After all, if you have gotten in the habit of doing something, and are experiencing increasing marginal returns, but are loving it and see no reason to stop, how is that addiction? The Chicago School definition addresses that by including a negative effect on welfare from “addictive stock”, defined as how much you have consumed in the past. This is the “running fast to just stand still” effect. An addicted smoker gets more benefit from each cigarette than someone else would because of past use, this story goes, but a lot of that benefit is used up just getting out of the hold, back to the neutral state she would be in if she had not smoked anytime recently. That is something that is theoretically measurable in people (though it is functionally nearly impossible — you have to ask them about well-being, and then figure out how to calibrate the answers and control for confounding), but not even theoretically measurable in mice.
But there seems to be one other candidate for addressing the “not wanting to be in the state” feature, the concept of second-order preferences.
(Actually, I can think of another candidate — it is a terrible concept, but so common it is worth mentioning: The “and also it is somehow bad” aspect can be defined in terms of the consumption hurting oneself. The problem with this is that every consumption choice involves hurting yourself, because you are giving up resources and other opportunities. So this always involves an implicit declarations that some harms that you inflict on yourself are fine, but others are not, which just plays into paternalists’ or collectivists’ pet notions. So it is common, but ugly, both politically and scientifically. Some will merely say you are not allowed to damage your body (because it belongs to the society, not to you), and thus addiction is anything with particular economic/behavioral properties that does any physical damage. Others will go further and effectively assert ownership over your productivity or even mental state, declaring that a choice to make those go badly is something you should be protected from — as in “sure smokeless tobacco produces trivial health risks, but it still means that someone is addicted to nicotine, and that is bad in itself, so we cannot allow that.” Even if you do not find that politically repugnant, it should be obvious that it is scientifically useless.)
I wrote about second-order preferences a few days ago (though I have yet to settle on whether it has a hyphen). I explained that they are a preference about preference, usually discussed in the context of a preference to have a different preference than you do not, as in “I want a cigarette, but I want to not want a cigarette.” This is another economic concept. Some would argue that it is better seen as a proto-economics concept because, though preferences are at the heart of economics, economics starts with preferences as a fixed point and works from there, and so anything that discusses the source of preferences or wanting to change them exists just before economics. Others disagree. Either way, it is in the neighborhood.
Mice may not be even be capable of having second order preferences, but people are tortured by them. A second order preference could make concrete the negative vibe in “addiction”: You experience increasing marginal returns, and so are best off consuming the good (i.e., you prefer to consume it), but you wish that things were different and you did not prefer to consume it. This is compatible with wishing you were not in the “running to stand still” hole, but that aspect is not necessary.
A defensible definition of “addiction” could have at its core “like it a lot, but often wish you could stop liking it”. I think this needs some element of varying preference, though maybe not specifically the increasing marginal returns. After all, if you like the experience/feeling/high and choose to do it, there must be something different about its appeal at the times you wish it did not appeal. Maybe it is just a moment before you start when you say “I shouldn’t do this”, but once you are doing it, that feeling is gone; that might be (rapidly) increasing marginal returns once you start. But it could be that you get satiated, like with a normal good, and that is when you wish you would not want more later.
One interesting feature of a definition that is partially built on second-order preferences might be that if you do not wish to not like the good, then you are not addicted, no matter what your behavior pattern is. To someone gathering objective data about you (and thus who cannot see your preference), you cannot be judged to be addicted until and unless you try to quit and fail, and even then you would need more data not sufficient (e.g., a single quit attempt could be an experiment to see if you might be happier without, and the discovery that the answer was no). But this kind of fits the colloquial usage. On the other hand, there seem to be some behaviors and preferences that somehow feel like they out to be addiction, if the word is to take its colloquial meaning, but there is no desire to stop liking them, so I am not sure it is wise to tie too tightly to the notion.
But it seems to play some role. For example, second-order preferences offer an easy explanation for why the NRT approach is so often tried unsuccessfully. I will try to get back to that point shortly.