Self-Interest is the Basis of Self-Denial

On Cooperative Equilibria and the Capacity for Commitment

A second pass at the themes in The Meta Level Doesn’t Justify Itself. I’ll roll the two together at some point in the indefinite future.

Imagine you need money, and somehow you find yourself sitting across from Warren Buffet pitching a new business venture.

From a purely self-interested perspective, your best scenario is to get the money, blow it all on raves and parties, and then run off to Cancun before he catches on.

Of course, if he knows you’re going to do this, he doesn’t give you the money at all. That’s your worst scenario.

Your second best scenario is to get the money, plow it into your business venture, and return it to Mr. Buffet later on along with a good portion of your returns. So in order to get to your second best, you have to be able to deny yourself the first best. This is hard because once you’ve achieved your second best, you have the option of jumping to your first-best.

“Lemme stop you right there…”

This is the problem of time-inconsistency. Kydland and Prescott formulated it in their Nobel-winning paper in terms of rules vs discretion. By not following your interests at every point in time, you can actually do better than if you did. This seems to be something similar to what Scott Alexander had in mind in distinguishing between the meta vs object levels of thought, even though the pitch is in terms of epistemic virtue rather than practical interest. Still, the common thread is that there are cases when it’s desirable not to act (or formulate opinions) in your own immediate interest.

I argued in the previous installment that the institutions of modernism – rationalism, capitalism, and liberalism – could all be formulated as rules over discretion, or meta over object level, and that modernism in general consists in the predilection for ever more abstract rules of this sort. Because problems of the Warren Buffet sort are pervasive impediments to social cooperation, these institutions were largely responsible for sparking the takeoff of Western civilization.

Both ways of formulating the problem are odd though. Presumably there’s some set of (unarticulated?) rules that direct discretionary action too; otherwise it’d be random action and we couldn’t call it purposeful at all. And not even an abstract rule like “pursue your own interests”, but a whole panoply of rule-bound heuristics and cognitive processes that turn environmental input into a course of action. In the end, everything is rules. So if we want to make a distinction like Kydland and Prescott do, it has to refer to the content of the directing rules rather than the existence of rules in general.

A more exact way to formulate the difference is in the capacity to commit to behaving in ways that might not be in your immediate self-interest. Returning to the Buffet example, it’s in your interest ex ante to commit, since this moves you from your worst option to your second best. But once you have the money, it’s no longer in your interest to have committed ex post, since this prevents you from moving to your first best. Commitment might consist in allowing a third party to coerce you in the event of defection (arbitration), or from having other-regarding preferences (where I value your interests even at my own expense – Bowles and Gintis argue that this is necessary for any social cooperation at all). Rationalism, capitalism, and liberalism were powerful because they were commitment technologies; they were legitimators of cooperative strategies. Similarly, following discretion = operating on the object level = failing to commit, all of which leave you in your least-preferred scenario.

The problem becomes more acute in situations where both parties have to cooperate. We can imagine a similar dilemma facing Mr. Buffet. In his first-best scenario, he convinces you to get to work, telling you he’ll give you the loan if the first results are solid. He then pretends to be dissatisfied with the results in order to avoid paying for it. Free labor for him!

Of course, if you think he’s going to do this, you won’t bother asking in the first place, which is his worst case scenario assuming you’re honest. So, like you, he has to commit to forswearing his first best in order to get his second best.

This is the essence of a prisoner’s dilemma: there are big gains to be made, but both parties have to commit ex ante to not follow their interests ex post in order to do better than the baseline. If either one fails to commit, neither one gets the benefits.1

With this in mind, I’m going to try to reformulate the problem of applying the meta-level to itself or trying to justify a rule with itself. If you value political neutrality between interests (a meta-level rule), this might look like being neutral between neutralists and anti-neutralists. If you value tolerance, this might look like tolerating the intolerant. If you value epistemic humility, this might look like being willing to countenance epistemic hubris. Something similar can be done to any rule that commits you to acting against your own ex post interest.

All of these cases describe committing to cooperate with a party you know won’t cooperate in return. You know Mr. Buffet won’t come through with the loan, but you come in with printouts of your credit score and swear on your mother’s grave that you’ve never defaulted on a loan. Mr. Buffet knows you’re going to run off with his money, but he brings you in and solemnly promises to appraise your progress with an impartial third party.

In cases like this where the other party hasn’t committed to cooperate, it’s not even in your ex ante interest to cooperate. Doing so will put you even worse than your previous worst-case scenario!

So to say that “rules don’t justify themselves”, or that “the existence of the meta level is itself an object level norm“, is just to say that commitment to cooperate has to be grounded in your interests. It has to be in your interests ex ante to commit to denying your interests ex post, which in general is only true when dealing with other committed cooperators.

Putting it this way makes it sound perfectly obvious. Of course it makes no sense to cooperate with defectors. And yet, few people think of the point of morals, rules, institutions, ideologies, and so forth being to foster cooperation.

The modern world being characterized by increasingly abstract rules, therefore, means that it is characterized by increasing capacity for commitment, and increasing scope for cooperation. This is a good thing – the Great Divergence, history’s most dramatic and sustained increase in the scope of social cooperation, would have been unthinkable without it. And yet, the fetishization of these abstract rules – the idea that they could justify themselves (modernism) without reference to some other end (cooperation) – resulted in disillusionment when they were shown to have failed to do so (postmodernism). The drive for abstract rules turns on itself, delegitimizes the project entirely, and reduces the existing normological order to ex post interest (deconstructionism). From such a standpoint, cooperation never pays.

At this point there are two main responses. Right-deconstruction says, “Stop cooperating! Never be a sucker!” – a workable survival strategy, but an impoverishing one that undoes everything that made the West successful and distinctive. Left-deconstruction says, “Cooperate, though the heavens fall!” – the ultimate commitment, but vulnerable to sudden extinction. And, also, defection in the higher-order game which requires the punishment of defectors. Unconditional cooperation cannot be an equilibrium, since now the other party has no reason not to defect.

The West is in a situation now where large swaths of the population are unconditional cooperators (at least in an important subset of games), masses of people from a society of defectors are washing in, and homegrown defectors are starting to build mass in response.

The sensible and stable strategy, of course, is conditional cooperation: cooperate only so long as it remains in your ex ante interests, and punish defectors to ensure that it does remain in your ex ante interests. The task of punishing defectors is itself a prisoner’s dilemma, or a commons problem, and the West is in a situation where the dominant ideology is that justice requires cooperating with defectors, which mostly precludes punishing them. And so antisocial behavior is subsidized on a gargantuan scale. Such is the result of taking justice as an end in itself, of assuming that it can justify itself, rather than justifying it by reference to cooperation as its end.

In closing, what might a society with healthy defection-punishment mechanisms look like, relative to the current political reality?

  1. More liberal use of the death penalty for violent crimes. A greater acceptance of type I errors and distaste for type II errors on the margin, considering most of the West is at a corner solution on that tradeoff now.
  2. Free migration for cooperators, but greater use and development of heuristics for determining whether potential entrants are cooperators or defectors. (Does skilled/unskilled approximate this closely enough?)
  3. Reversing hyper-inclusionism, which prevents the accumulation of common knowledge in organizations and limits the scope for cooperation. (Does this entail something like Keiretsu?)
  4. Devolution of political power to more local (subnational) bodies, for the same reason.
  5. (What else?)

The upshot is the same in each case: promote cooperation by efficiently weeding out defectors. Sustainable cooperation cannot be an end in itself; it must be grounded in (but not limited to) a more narrowly conceived interest.2 If this proves unachievable, the alternative is the end of impersonal cooperation and the extended order.


  1. Parties can be fooled, but not permanently. “Fool me once…”
  2. Though this need not be conscious; that interest could just as well be evolutionary fitness, advanced by cooperation which is an end in itself on a conscious level.


CooperationCultureNormsEdward PrescottFinn KydlandHerbert GintisSam Bowles


Facebook Twitter Reddit Threads


  • 1

    Jim Caton

    Aug 13, 2016 at 17:50 | Reply

    Disempowering or banishing defectors seems adequate for maintaining the integrity of the group. Why do you think the death penalty is appropriate?

    • 1.1

      Cameron Harwick

      Aug 13, 2016 at 17:57

      Partly symbolic because death penalty abolitionism is symptomatic of a more general unwillingness to punish defectors (which, I should note, is itself defection in a higher-level game). And partly substantiative because defectors can continue to cause problems (banishment), cost a great deal of resources (imprisonment), and/or reproduce (problematic if the propensity to cooperate is at all heritable).

    • 1.2

      Jim Caton

      Aug 17, 2016 at 15:25

      Their ability to cause problems is dependent on asymmetry in information, which is not as much of a problem today. The cost of imprisonment is a problem, but would be relieved by banishment. Any reference concerning reproduction is speculative and discounts free will the ability of individuals to overcome these problems. The possibility of banishment is itself a substantial punishment. The life-role of that individual is obliterated.

    • 1.3

      Cameron Harwick

      Aug 17, 2016 at 16:52

      On asymmetry of information: what’s the prototypical interaction you have in mind here?

      On banishment: seems like a beggar-thy-neighbor situation. Where do you banish them to? Unless it’s the woods or something, you haven’t solved the problem, you’ve just pushed it off on someone else. And in the case of sociopaths, these are people who aren’t that attached to the local group in the first place, so it wouldn’t be much of a punishment anyway.

      Maybe there are people who are defectors in some cultural contexts and cooperators in others, in which case repatriation would be a sensible strategy too.

      On reproduction: not so speculative. Psychopathy is between 66 and 90% heritable and conscientiousness about 40%. That’s huge if a community is unable to bring selective pressure to bear on those traits.

      On free will: choice springs from somewhere, and many behaviors and proclivities are ingrained in the structure of the brain. Does that make them unfree? Not on any definition of “freedom” that could be conceivably realized.

  • 2

    Brad Kells

    Sep 22, 2016 at 15:01 | Reply

    Good thinking here. A few thoughts:

    1) I’m not sure approximating leftists as unconditional cooperators makes sense. The “unconditionally cooperating” left cooperates with people from a foreign society of defectors, but not with the local society of defectors.
    1a) Does the fact that one group uses cooperative language actually make them more cooperative?
    1b) I still think Scott Alexander’s theory of tribes explains this more effectively.

    2) If at one point society more closely hewed to your heuristic, that is, cooperated when in their best interest and not otherwise, why was that the case? What was the justification?

    3) Utility functions are far more than monetary payoffs. I think part of the issue here is that the moral payoff from unconditional cooperation may have grown over time? People are willing to be betrayed, because they are unwilling to take the huge mental punishment from type one errors (stereotyping, immigration restrictions, etc).

    I think my problem here can be summarized thusly: if people have utility functions that are themselves functions of wealth, and that as wealth increases the marginal benefit of an additional dollar declines relative to the marginal benefit of fewer type 1 errors about others, then a) doesn’t this explain all evidence you are wondering about, and b) mean that your solutions would be harmful to net social utility?

    NB: In the above hypothesis, local defectors (right-deconstructionists) can be explained as simply individuals with lower than average utility loss from Type 1 errors. These fellows would only have an incentive to emerge after the first group gains power, and are also utility maximizing (in this case, maximizing the rest of their utility function by accepting Type 1 errors). Here your upshot holds.

    • 2.1

      Brad Kells

      Sep 22, 2016 at 15:49

      Further: to the meta-object problem, you write:

      “Such is the result of taking justice as an end in itself, of assuming that it can justify itself, rather than justifying it by reference to cooperation as its end.”

      Thus I take you to be arguing we should justify justice with respect to cooperation: that is, something is just because it allows these good things to happen, it allows cooperation and some degree of trust between actors (necessary for a stable society).

      But as I said last time, doesn’t this just move the game up a level of cooperation? You want individuals to respect norms that because create a good society. You make the meta-level object level, or at least change which meta-level we refer to.

      But your definition faces the same problem theirs does: your utility function is subjective.

      I agree that “cooperate regardless of what your opponent does” is a failed PD strategy. You can tell someone that, but you can’t change their utility being positive in not-making-Type-1-errors. You can educate them, which I think is the best strategy, that their strategy doesn’t effectively optimize their utility function long-term. But you can’t claim, as far as I can tell, that their definition of justice is falsely rooted.

      “…commitment to cooperate has to be grounded in your interests…”

      Yes. But which interests are those? Loyalty to tribe and to ideology can pay off as strongly as actually getting that loan from Warren.

Leave a Reply

More Content