Monday, January 30, 2012
The Ambiguity of Conscience and the Prisoners' Dilemma
In reading Lynn Stout's "Cultivating Conscience," I went back and forth as to whether or not I felt she was treating law and economics as a bit of a straw horse. There is plenty within the field, especially with the growth of behavioral law and economics, that is sympathetic to her position. But I do think there is a resistance to her position that goes deep into the DNA of economics, and after reading the book I felt both that her big point (people aren't as selfish as economists normally assume, and we need to consider how the law can promote a lack of selfishness) is worth hammering home and that she has many good specific ideas and examples. But to keep things interesting, I will focus here on one big point to which Stout pays too little attention.
With occasional caveats, Stout treats conscience ("an internal force that inspires unselfish, prosocial behavior") as unambiguously good. In reality, there is often much ambiguity, and it really matters. We can see that right in the classic game theory example of prosocial behavior, the Prisoners' Dilemma. The story, of course, is that two criminals have been caught, and prosecutors are trying to get each one to squeal on the other. "Cooperation" is defined as not squealing, and from the point of view of the prisoners, both not squealing is the optimal outcome. But from society's point of view, that's not right at all. We want them to squeal, in order to better punish the crime. Honor among thieves is great for thieves, but not for society as a whole.
More generally, the question is prosocial behavior in whose favor? What if there are conflicting possible groups towards whom one might be loyal? The question is ubiquitous. Within my and Stout's core field, corporate law, this comes through in the question of to whom officers and directors owe their fiduciary duties. I think that Stout adds great value by putting a focus on how fiduciary duty affects social norms. But consider the standard formulation that directors should advance the interest of shareholders. Stout is critical of this where the interests of shareholders diverge from broader social interests, and she makes some great points. But, getting directors to advance shareholder interests is itself prosocial behavior compared with them advancing their own personal interests, and that is the core object of fiduciary duty. How can and should corporate law best help shape the norms that guide directors? If we formulate those norms as much broader than advancing shareholder interests, will the resulting vagueness undercut the law's strength?
Stout's framework, and her focus on how the law affects norms, is the right starting point for asking questions like this. But that framework needs to be much more aware of the possibility that individuals may face claims from competing social groups and interests, and which ones the law should encourage them to follow can be a hard question.
Posted by Brett McDonnell on January 30, 2012 at 02:23 PM | Permalink
TrackBack URL for this entry:
Listed below are links to weblogs that reference The Ambiguity of Conscience and the Prisoners' Dilemma:
I think Stout's book, especially the anti-selfishness stuff, needs a good review by Ayn Rand.
Posted by: Jimbino | Jan 30, 2012 2:48:29 PM
As an aside, and speaking as an economist, saying that "people aren't as selfish as economists normally assume" is logically equivalent to saying that "people aren't usually as psychotic as psychologists normally assume": it mistakes a special case of a discipline's assumption for the general case.
What economists normally assume is self-interested behavior, defined to mean that we assume people pursue their interests as the people believe them to be. Since selfish people are people, too, it is accurate to say that economists normally assume that selfish people will pursue their interests self-interestedly.
But it does not follow that we normally assume that people in general are selfish. There is nothing in economic methodology--much less in "the DNA of economics"!--that assumes or even suggests that people perceive their interests to be what non-economists mean when they use the word "selfish". The failure to understand the distinction between selfishness and pursuit of one's interests is one of the most vexing (and, unfortunately, common) errors that critics of economics make. It is easy to write down models that incorporate altruism and/or concern for the well-being of others, and economists have been doing so for a long time.
I don't know whether it's Stout or McDonnell who's making the mistake here, as I haven't read Stout's book. But I don't think someone who makes this mistake is likely to have much success in communicating with or convincing people who do understand economic methodology.
Posted by: jonah gelbach | Jan 30, 2012 3:17:40 PM
Stout does discuss that, and I am aware of it too (I have a Ph.D. in economics). But I think Stout's answer is pretty persuasive. Although as a general formulation objective functions are indeed quite broad, in practice the usual assumption is selfishness. And there is good reason for that usual assumption: an overly broad formulation risks reducing the idea of rational utility maximization to tautology. You can explain just about any sort of behavior if you assume the right objectives.
Now, there are various alternatives in between selfishness and tautology, and Jonah is right that for a long time various economists have created models in that space in between. When I said that there is plenty within economics, and law and economics, that is not inconsistent with Stout's position, that is what I had in mind. Still, creating such models and making them really testable isn't easy. And I do think the usual starting point is selfishness; there are a large number of exceptions, but selfishness remains the default assumption for your standard concrete economic model. And in the end I think that Stout is right: that default assumption matters, even though economists quite frequently move away from it.
Posted by: Brett McDonnell | Jan 30, 2012 3:38:46 PM
straw man, or stalking horse?
On a more constructive note, I think your general criticism is good. How can anyone expect to make a coherent, rigorous argument was such ridiculously capacious and manipulable concepts as "prosocial" or "unselfish"?
Your use of the prisoner's dilemma is awkward though. The problem is not that, if bad people cooperate, they can more efficiently perform bad acts. The prisoners in the dilemma constitute the entire relevant society, and cooperating maximizes collective welfare, so dwelling on their status as "theives" takes the construct too literally.
What the PD illustrates is that even if people are really "prosocial" like Stout wants, individual prosocial maximizers may not maximize collective welfare. So, if I am a crusader for rescuing kittens, and you are a crusader for rescuing puppies, rational choice teaches us that collective action problems created in the PD context where these noble goals conflict will likely mean that the socially optimal number of orphaned animals is not saved. That's the irony Stout misses by ignoring basic law and economics and preferring simple exhortations that we should all be less "selfish."
Posted by: Matt | Jan 30, 2012 6:10:34 PM
I think when you write that "selfishness remains the default assumption for your standard concrete economic model", you are conflating parsimony -- model the things that are most relevant to the topic you are studying -- with "default assumption".
Posted by: jonah gelbach | Jan 31, 2012 2:35:42 PM
But what are "the things that are most relevant to the topic you are studying"? One way of putting Stout's objection is that economists are far too prone to think that they shouldn't assume conscience. They tend to think that if self-interest narrowly construed is adequate to produce a model that explains what they are studying, parsimony prefers assuming narrow self-interest rather than conscience. And indeed, once you add in phenomena like repeated interactions and reputation, you really can explain most behavior without reference to conscience.
But Stout argues that such a notion of parsimony leads us to ignore conscience even where it is a major presence, and that in turn leads us to be too cynical. And indeed, there is evidence that economists are more likely to assume selfishness in others, and to be more selfish themselves (whether this is a result of economics education or a selection effect, or both, is a matter of some dispute). This effect is the core normative objection that Stout has to law and economics, although as an explanatory matter she also thinks that much experimental gaming evidence has shown that we really do need to assume conscience to adequately explain much behavior.
Posted by: Brett McDonnell | Jan 31, 2012 3:02:23 PM
I thought I posted a reply to your thoughtful reply to me, but weirdly it doesn't seem to have shown up, so maybe I screwed up in posting. So here's my quick summary of what I wrote (apologies if I am duplicating):
1. I think you (and evidently Stout) are missing the point of much economic theorizing. Very frequently the point isn't to explain everything, or even to explain the thing in question in the most complete way. Rather it's frequently to understand the types of effects that operate in a given context.
Here's a perfect example (though there is no shortage of others): Akerlof's lemons model. Suppose Akerlof allowed used car sellers to have a conscience. Either the conscience would be strong enough to eliminate the lemons problem or it wouldn't. Do you believe it's strong enough to eliminate the lemons problem in ALL transactions? If not, then you are back to Akerlof's model, in which he ignores conscience. (If so...have I got a car for you!) Adding conscience to that model would needlessly waste time and destroy clarity. If Stout's "core normative objection" is to this sort of modeling, then I am unconvinced, to say the least.
2. You write that "once you add in phenomena like repeated interactions and reputation, you really can explain most behavior without reference to conscience." And then you refer to Stout's belief that "much experimental gaming evidence has shown that we really do need to assume conscience to adequately explain much behavior." You can't both be right!
Posted by: jonah gelbach | Jan 31, 2012 4:45:46 PM
It seems to me that whether something is prosocial depends on your definition of the good society. On many issues, I suspect Brett and I would have very different definitions of the good society. He might think the good society promotes reproductive choice, while I might think the good society respects all lives whether they have been born yet or not. From this perspective, when you or I call something "prosocial" all we're really doing is using a 50-cent word that really means "pro how I think the world should be run." And I doubt that changes if you change the label to conscience or selflessness or what have you. At some point, all of the discussion assumes a single set of normative priors when there may be many such sets.
Posted by: Steve Bainbridge | Feb 1, 2012 8:14:15 PM
Jonah (responding to your last comment post): two thoughtful points. On the first, I'm with you much of the way. As I said originally, I too had a reaction that in many ways Stout is not fully fair in her characterization of economics. But still, don't you think that what kinds of things one wants to explain, and what sorts of arguments, and underlying assumptions, are seen as offering the best, simplest sorts of explanations, tells you something about a field, and those in it? Even if the modelers don't believe in the realism of their assumptions in the first place (and I think they often do), they come to believe in them when they get made repeatedly across most models. It's not inevitable that one need use the narrow self-interest assumption as often as economists do, and scholars in other areas of social science do not make that assumption nearly as frequently.
On your second point, touche, but here's what I was trying to say. For most real world phenomena, I think that narrow self-interest will usually be able to explain most behavior once one includes repeated interactions, social context, and reputation. To really separate out conscience, one needs the artificial setting of experimental games. There, many experiments do suggest prosocial behavior even where all possible selfish incentives supporting the same behavior seem to have been removed. I'm not quite sure how much weight to give such evidence.
To Steve: you might be surprised by my views on your particular example, especially if one is discussing norms rather than laws. But on the general point, I agree completely. I don't think that's an argument for assuming narrow self-interest as economists do too frequently, and I don't take you to be making that point. But it is an argument that the implications of "pro-social" behavior are much more complicated than Stout usually assumes--which fits well with the point I was trying to make.
Posted by: Brett McDonnell | Feb 1, 2012 10:17:15 PM
I think your question,
"don't you think that what kinds of things one wants to explain, and what sorts of arguments, and underlying assumptions, are seen as offering the best, simplest sorts of explanations, tells you something about a field, and those in it?"
is an interesting one. And yes, I do think that what kinds of things one wants to explain tells you something about a field. But economists do engage issues like conscience, or altruism, and so on, when there's a reason to--when they are interested in explaining phenomena where those issues are potentially important in the sense of not just being modeling detritus. But I still don't think that there's any real info in the observation that economists don't hurl a conscience argument into every objective function.
One last point on the question of whether "modelers . . . believe in the realism of their assumptions in the first place". I think you are still kind of missing my point. You read the decision not to include, say, conscience in a model as an assumption that there is no conscience. I think that the real assumption is just that conscience isn't that important to the issue at hand. I am quite sure that all economists believe the IRS exists, and yet it generally doesn't appear in economic models that don't involve tax issues. That's not because economists don't believe in the IRS! Actually, this brings up an impish observation. Most models don't involve drug abuse, or crime, or violence. Should we believe that economists are uncommonly anti- these things, too? :-).
Posted by: jonah gelbach | Feb 2, 2012 2:50:03 AM