Wednesday, December 31, 2014

[E]valuation is nontrivial

You know how when you're in a debate with a smart person, and the debate wants to fragment -- you need some kind of tree structure to keep track of all the threads that the two of you have brought up, because some of them are irrelevant to the main point but still intriguing, and some of them neither of you are sure whether they'll be crucial or just trivial, and sometime what was irrelevant turns into its own fun conversation/debate?

Except usually you can't do all of this, because the debate is taking place in a linear format like blog (or, worse, Facebook) comment threads?

I had a moment feeling a little like that today when reading Massimo Pigliucci's report from the American Philosophical Association meeting in Philadelphia where he relates sitting through a panel stocked with followers of Ayn Rand. In general, this is an exercise not worth the investment of time, since no one except Objectivists takes Objectivist philosophy seriously. (Not least because Objectivist philosophers don't take any non-Objectivist philosophy seriously, and it's pretty bootless to try to have an academic discourse that way.)

Massimo posted a running commentary on the panel, which of course included a lot of little points without a fully ramified response to any of them. There was one in particular, however, that I think perfectly exemplifies the shallow thinking of Randian philosophers on both morals and economics (which really aren't separate domains for them, or really for anyone, I suppose).

Yaron Brook, as reported by Massimo:
Rand stated the obvious when she claimed that you cannot put a gun to an engineer and tell him that from now on 2+2=5. A little less obvious is the similarity she draws between that case and the idea that one cannot put a (governmental) committee in charge of telling a company the value or appropriateness of a new drug. The metaphorical gun in the latter case, of course, is the regulatory force of government, and for Rand it is far more devastating than the one pointed to the head of our unfortunate engineer.
So let's think about this claim, and this comparison. I think everyone is in agreement that (a) 2 + 2 is not equal to 5, and (b) it would be unjust for the state to exercise its coercive power to coerce anyone to assert that it was. More generally, we can say that to the extent that an assertion is one of objective and verifiable/falsifiable fact, it would be unjust for the state to coerce a contradictory assertion from someone.

(Leave aside for today the problematic question of what, fundamentally, objectivity means. Posit for the sake of argument that some facts are objective; we won't be dealing with any real problem cases today. OK, one problem case, but I'll be arguing that it's not really a problem case at all.)

Of course, the bolded caveat is doing a lot of heavy lifting here. Mathematics is a domain with a lot of "objectivity" to it -- given premises and a putative conclusion and fixed rules of inference, the conclusion either follows from the premises or it doesn't (though there is no algorithm to determine which). When we step into statements which are not analytically true or false, there is quite a bit more gray.

Before I get to the specific example in Brook's talk, let's think about a slightly less complex example taken from the medical field. Suppose the state (or an agent of the state, such as the Standards and Practices for Care of the medical licensing board) were to order physicians to assert that antibiotics do not kill bacteria: would this be tyranny or a just exercise of power?

You've probably already guessed that the answer is "it depends". The ambiguity stems first from the implicit quantifiers in the statement "antibiotics do not kill bacteria". There are two such implicit quantifiers, and it is not at all obvious what they should be interpreted to mean. If "no antibiotics kill any bacteria" is meant, then the assertion is simply ludicrously false (and has no business being mandated by the state); however, if "all antibiotics may fail to kill some bacteria" is meant, then not only is the assertion true, but serves a valuable public health service.

But more than the simple ambiguity of the assertion itself is the intention of the state in my hypothetical: such a directive would have a very different character as a matter of justice if it came out of the blue, than if it came accompanied with information about what a physician should be attempting to get the patient to positively understand (e.g. the danger of antibiotic-resistant bacteria and how to avoid creating them).

Now, killing bacteria is an observable, objective fact, and we can clarify the statement "antibiotics do not kill bacteria" into a precise and objectively true assertion (though I personally think the statement as it stands should not be mandated), something along the lines of "antibiotics do not cure many diseases, such as the common cold and flu; and there exist versions of some bacteria which resist being killed by basically all known antibiotics". This is neither here nor there, of course: this example is merely to point out that even when there is an objective fact of the matter, the matter is not so clear-cut as in the case of math.

(A minor aside: I have, on more than one occasion, been asked a great question in a math class, in a context where answering the question as it deserved to be answered would have dragged the class away from what needed to be learned that day or that semester. In these instances, I have been known to literally say "Now, I'm going to lie a bit, because your question is way more complicated than this -- but here's what's going on" and proceed to grossly oversimplify, and promise a real answer in whatever course deals with that topic. I don't feel at all guilty about these "lies", especially because they're acknowledged.)

OK, so let's turn now to Brook's actual hypothetical: that of a pharmaceutical company which has a drug it wants to sell, and which the state wants to interfere in the selling of (either by restricting the diagnoses which the drug may be applied to, or the price at which it can be sold). Brook's thesis is, as near as we can tell, that the "value and appropriateness" of the drug are just as much a matter of objective fact as the value of 2+2, and the pharmaceutical company knows that objective fact while the state should not be assumed to; it is hence unjust for the state to constrain the company's recommendations about use of their drug.

As to appropriateness: in the absence of a publicly audited testing regimen, under which the state can coerce the company to keep the drug off the market if its effects are insufficient or its side-effects out-of-bounds, why would we suspect that the company has any idea what maladies their drug is appropriate for? Under a purely laissez-faire conception of rights and duties, the company's only duty is to make money for its owners -- it has no obligation to the health of its customers at all (leaving the invisible hand of the market to aggregate data about the value of the drug from satisfied and dissatisfied consumers). Even under a more enlightened libertarianism, one in which product claims must be true (because fraud is a form of aggression), why should we expect the company to have set itself an adequate testing bar to jump over, to filter out placebos and false positives and find unfortunate side effects? I mean, maybe they're morally obligated to report side effects if they find them, but do they have an affirmative duty, in this conception of morals/ethics, to seek those side effects out? And if so, how long do they have to search before they conclude they've found them all?

The main point is the naiveté of trying to claim that either (a) from a consequentialist perspective, the way to maximize human good is for actors to pursue their own self-defined ends without any effort to balance any ends which may be in competition, or (b) from a deontological perspective, if an actor ought to behave a certain way (and acting otherwise harms others), then it is immoral for the state to coerce them to act that way.

But we're not done! For what really made Brook's statement jump off the page to me is the assumption that the "value" of the drug was just as much a matter of objective fact as the efficacy of the drug.

I mean, what is the value of the drug supposed to mean? Well, for starters, there are two different, incompatible units which it makes sense to quantify such a value in: dollars (or whatever currency medium of exchange we want to use), or person-years of added life expectancy! (For the moment, we'll assume that the drug keeps one from dying, rather than raising one's quality of life from a low but nonlethal level to a higher level.)

Not only that, but let's think about these measures definitionally cannot be computed independently of each other. Let's see why not:

First, let's think about how to compute the monetary value of the drug to the company. Let's take a very simple one-timepoint model, with a demand-curve for the drug. What the company would love to do, to maximize its profits, is to sell to each consumer at the maximum price that consumer is willing to pay, provided that price exceeds the cost to the company involved in producing that dose. Of course, this strategy does not work in the real world; instead, the company finds a point on the demand-curve which maximizes profit, and charges all consumers the same price. Naturally, at this price, some potential consumers of the drug will be priced out of its benefits and will instead die, so the aggregate person-years of added life-expectancy will be lower than they would have been had the drug been given away freely (or on a pay-what-you-will model).

Of course, it is completely banal to point out that a pharmaceutical company's profits are in competition with maximizing the health effects of its own products. However, the situation is even worse: remember how we assumed the existence of a demand curve? Where do those come from? A demand curve is an abstraction meant to represent the resources available to all the agents in an economy, as well as the competing goods among which they can allot those resources. In other words: the (monetary) value of the drug is a function of the demand curve, and the demand curve is a function of the rest of the economic environment. So it makes no sense to speak of "the value" in monetary terms of any good, as if the resources (i.e. money) which people would be willing to exchange for it remained constant under wildly different assumptions (e.g. what if there were no state-issued money at all? what if each person had to travel many miles to fetch the drug rather than having it delivered to the hospital where they were in treatment?...).

And yet, Brook and the rest of the panel seem oblivious to this notion, that "the value" of a good may not be in any way an "objective fact", or at least not when denominated in monetary units. "Value", even as narrowly construed as "amount of money people are willing to pay me in exchange", is highly context-dependent. The actions of the state can change the demand-curve, which changes the value; either by mandating that the price level be set above (or more likely, below) where the company would "otherwise" choose, or providing other forms of incentives which people's preferences and allocations will respond to.

I guess what I'm saying is that there is no "state of nature" in which we find demand curves -- and if there were, I don't think I'd want to live there anyway.

No comments:

Post a Comment