Yes! A Real Newcomb Case!

It’s time to reopen this can of worms. Michigan researchers have potentially discovered a gene responsible for both smoking and cancer:

“The gene is not the only element responsible for regular smoking, but it does signal a risk factor for nicotine dependency and cancer.”

Ann Arbor News reports on it here.

A more in depth article is here along with a citation to the research.

Thoughts?

Advertisements

10 Responses to Yes! A Real Newcomb Case!

  1. Dustin says:

    Sorry to be a downer, but even if their conclusion is right, this isn’t a Newcomb case. If this were a Newcomb case, cancer would have to be (sufficiently) causally independent of smoking. The article doesn’t support that claim. As far as what the article says, it could be that smoking still (sufficiently) causally increases your chances of cancer. In fact, it could be that the gene causally increases your chances of cancer *via* causally increasing your chances of smoking.

  2. That’s right, Dustin. Though, it’s compatible with this evidence that smoking is in fact a newcomb case, also. It simply suggests that the types of situations that are considered in the newcomb cases, which are often rejected as completely implausible, are more plausibly real.

  3. Dustin says:

    “Though, it’s compatible with this evidence that smoking is in fact a newcomb case, also.”

    That’s always been the case.

    “It simply suggests that the types of situations that are considered in the newcomb cases… are more plausibly real.”

    How so? As far as I can tell, all we have is evidence that there is some factor X (namely, a gene) that makes both smoking and cancer more likely. But we’ve always had reason to believe that–that is, there’s always been reason to believe that there is some factor (perhaps sociological, perhaps genetic, perhaps something else) that makes smoking more likely and thus (thus!) makes cancer more likely. What we for what you say to be true is first and foremost (A) evidence that this gene makes cancer more likely independently of making smoking more likely and (B) evidence that smoking itself does not sufficiently increase the chances of cancer. Perhaps we’ll discover (A), but I would be shocked (shocked!) if we were to discover (B).

    Anyway, who rejects Newcomb cases because they are implausible? That’s just silly.

  4. Let’s assume that you’re correct that in order for this to be a Newcomb case, it must satisfy both (A) and (B).

    If you take the news article to be correct, then it suggests that (A) holds. It claims, “The gene … does signal a risk factor for nicotine dependency and cancer.” I take that to mean that it makes cancer more likely independently of smoking (although an alternative reading of it would not be completely implausible). Take, for example,

    (1) Drunk driving increases your likelihood of being in a car accident and suffering accident-related injuries.

    (2) Drunk driving increases your likelihood of being in a car accident.

    In this case, (1) contains more information than what is contained in (2). The reason why is that (1) also implies that Pr( accident-related injuries | accident and drinking ) > Pr( accident-related injuries | accident and not drinking ), which is not contained in (2). Similarly, it appears that the article claims that Pr( cancer | gene and smoking ) > Pr( cancer | gene and not smoking ). Hence, the example provided satisfies your condition (A).

    Clearly, (B) is not suggested by this article. So, you’re right this study does not sufficiently show that this is a Newcomb case under your definition.

    It’s not immediately obvious to me why (B) is a necessary condition for this to be a Newcomb case. I think that the condition you’d really want is (B’) that smoking does not *cause* cancer. Maybe I’m misunderstanding the problem, but if assumptions of the smoking Newcomb problem are satisfied, then there still is a high correlation between smoking and cancer (that is, Pr ( cancer | smoking ) >> Pr ( cancer | not smoking) ), but not because smoking *causes* cancer rather because the gene causes both of them. So, your condition (B) would not be satisfied by a Newcomb case.

    This article does not suggest (B’), so in that sense, it is not a full Newcomb case. I would have to think about it more, but prima facie, it seems like you’d be able to get a Newcomb-like decision paradox though only given (A). I’ll have to return to that line of thought later.

  5. Dustin says:

    Hi again, Daniel. Your reading of “The gene … does signal a risk factor for nicotine dependency and cancer” is of course possible, but not very plausible, mostly because it is not very charitable. If that’s the reading the author intends, then his conclusion goes well beyond any evidence presented in the article. My reading has the advantage of having the author’s conclusion supported by the evidence he presents. Moreover, note that he makes a point to say this: “Other researchers have already linked variations in the same genetic region to smokers’ level of dependence on nicotine, to the number of cigarettes smoked per day and to a far higher risk of lung cancer — the ultimate outcome of a lifetime of smoking.”

    As far as (B`) goes, no, what I want is (B). By “smoking increases the chance of cancer” I mean “cancer is more causally likely given smoking than given not smoking”. If you change your “smoking causes cancer” in (B`) to something like “smoking statistically causes cancer” or something like that, then they would be equivalent, at least as I intend them. In any case, I don’t think we really disagree on this one. Probably just one of those subtle communication problems that arises in oh so many blog conversations.

  6. Hey there again!

    Umm … I am still not inclined to think that my reading of that sentence is implausible. Here is another section of the article that seems to support that position:

    “He also notes that, this year, three papers published independently of one another demonstrated that variations in the same gene, and related genes, greatly increase the risk of lung cancer.

    Taking into account its links to increased liking of initial smoking, stronger likelihood of getting addicted to nicotine, and greater probability of developing lung cancer, this genetic variant may well constitute a “triple whammy” for smoking-related disease, he says.”

    I take this section to suggest that the gene makes lung cancer more likely independently of smoking, but in the case where the person bearing the gene does smoke, it increases his risk even more.

    Also, with respect to (B) vs. (B’), what I was trying to claim is that in the Smoking Newcomb case, any statistical/probabilistic analysis should give you the result that smoking statically causes cancer. This is the case because in the perfect Smoking Newcomb case, most people who smoke have the gene, and hence, have cancer (of the increased likelihood thereof). On the other hand, what separates the Newcomb case in this situation is that in the Newcomb case, smoking does not actually cause cancer (even though it is a statically correlated with it). I am inclined to believe we’re on the same page about this, too though.

    ((Meta comment: Also, I just noticed that I can edit *other people’s* comments on this post. What’s the deal with that? I could just make you say whatever I want apparently … ))

  7. dtlocke says:

    OK.

  8. dtlocke says:

    FWIW, the authors of the paper published this week write:

    “Taking into account other recent studies, the findings suggest that early smoking experiences may mediate nicotine dependence, involving not only alpha-5 subunits but possibly also beta-2 subunits as well as alpha-6 and beta-3 subunits. Entrainment of nicotine dependence via enhanced sensitivity to nicotine also provides a *plausible mechanism* for explaining findings from three closely related investigations showing that risk of smoking and lung cancer is associated with a group of genes in chromosome 15 coding for alpha-3, alpha-5 and beta-4 nicotine receptor subunits.” (emphasis added)

    Not that I understand all that.

    —–;—@ <—pretty rose

  9. dtlocke says:

    Yo, use that “edit comment” feature to make my comment look pretty, would ya.

  10. Yea, it’s beyond me at this point.

%d bloggers like this: