Discretion Is Necessary For Commitment Power
Political Economy4
Excerpted from "Inside and Outside Perspectives on Legitimacy"

Discretion Is Necessary For Commitment Power

The usual supposition in economics is that rules are generally preferable to administrative discretion, in that the latter often precludes credible commitments, and introduces a social dilemma that an enforceable rule could solve (Simons 1936; Kydland and Prescott 1977; Root 1989).

This logic holds to the extent that one can rely on exogenous enforcement somewhere in the logical chain, whether from the state or from preexisting norms. In the case of monetary policy, for example, this can be a valid assumption, as the central bank is generally already embedded in an enforcement apparatus. Suppose, therefore, that a society, desiring to pre-commit itself to cooperation now and for eternity, delegates the enforcement of cooperation to a disinterested and omniscient robot. There is no necessity of punishing the robot, as it necessarily obeys its programming. Is this sufficient to close the incentive gap?

The problem is this: if it is true that there always exists some opportunity for profitable (in terms of the objective payoffs) defection from a society’s norms, then it will also be true that any conceivable set of pre-announced and rigidly followed rules will be vulnerable to exploitation. It follows, then, that there must always be an irreducibly discretionary element to a society’s mechanisms for protecting itself from defection.

A course of action is ‘discretionary’ if it is decided on the spot rather than according to pre-announced rules. There are two very different things this can entail. The traditional economics of commitment takes the lack of precommitment to entail rationally maximizing behavior, i.e. the sort of behavior that – per section 2 above – would make society impossible if pursued consistently. But refusal to commit to an explicit rule does not necessarily entail rational maximizing for agents with altruistic preferences. Rather, a more basic feature of this refusal is the tacitness of the rules used to generate action.

Taking discretion to mean tacit decision rules flips the logic of rules versus discretion on its head. Indeed, rational maximizing – being in principle fully explicable – should be considered another variety of rule-bound behavior. And in this sense, it is the very lack of a fully articulable decision rule that generates commitment power in certain contexts. Some tacit and inarticulable “gut sense” of being exploited is necessary for identifying and punishing defectors, lest a rigid or preannounced rule set be invaded by a particular form of defector – and indeed the necessity of such a flexible “gut sense” seems to have been a significant driver of the development of human intelligence (Cosmides et al. 2010). Hayek’s (1952: 192) dictum that the mind can never fully explain itself is, therefore, more than simply a principled limit to self-knowledge; it is also a precondition of social behavior.

Robocop, therefore, is not a viable method of precommiting to open-ended cooperation, for the same reason that complete contracts are impossible. Any mechanical enforcement of preannounced or pre-programmed rules, however nuanced those rules may be, must eventually fall to exploitation without tacit human judgment as a backstop. Perception of a situation as an irreducible whole – what Hayek (1952) called Gestalt perception – is an essential aspect of human motivation and social organization, and one that cannot in principle be replicated programmatically.1

One straightforward policy implication is that the push toward uniformity in judicial sentencing, for example with mandatory minimum sentencing guidelines, is likely to ossify the ability of bureaucratized societies to maintain cooperation, at least in the criminal justice arena. While the logic of the law may be of necessity fully articulated, the rules governing its application in particular cases is, and must remain, tacit. Vague and imprecise legal concepts such as negligence, reasonableness, and so on, are important bulwarks against novel forms of defection that arise in response to existing complexes of articulated rules. Any functional legal system must rely to some extent on the legitimacy of the decision rule Justice Potter Stewart used to delineate the category of pornography in Jacobellis v. Ohio: “I know it when I see it”.

Second, the recent development of programmatic contract enforcement and trustless exchange with blockchain technology cannot substitute for judges and juries any more than an explicitly programmed Robocop can enforce cooperation upon a community. These developments can be extremely valuable, especially to the extent that they can relieve courts of the burden of cases where the mechanical application of a rule is sufficient. This may indeed be many cases, but it can in principle never be all cases. In this sense smart contracts, DAOs, and other programmatic contracts are a complement to, and not a substitute for, traditional contracts enforced by the judgment of a human mediator.

Finally, the importance of Gestalt perception to the maintenance of cooperation indicates why sacredness is such a central feature of human experience. Something sacred resists analysis, resists being broken down into its constituent concepts; it must be apprehended as a whole. Indeed, the attempt to analyze sacred concepts frequently provokes offense. This reaction is especially familiar to economists, who are in the business of pointing out tradeoffs. It is, of course, valid to take an outside perspective on sacred values and point out that – for example – a human life can have a concrete opportunity cost that may not be judged worthwhile to bear, hence the EPA’s much-maligned “value of a statistical life”. At the same time, economists ought to appreciate the functional role of sacred values in human cooperation. Particular sacred values may – and often should – be criticized and analyzed, but sacredness itself cannot be dispensed with or rejected as irrational.


  1. The opaqueness of the processes behind “deep learning” neural network AIs (Knight 2017) – despite typically being seen as a problem (e.g. Park et al. 2017) – could in principle suffice. Of course, given that the legitimacy of punishment is the key constraint, an opaque algorithm is not likely to command much assent.


CooperationEpistemologyLawEdward PrescottF.A. HayekFinn Kydland


Facebook Twitter Reddit StumbleUpon


  • 1


    May 24, 2018 at 12:36 | Reply

    Interesting. I guess in a way compares to letter if the law vs. Intent?

    • 1.1

      Cameron Harwick

      May 24, 2018 at 12:43

      I like that example, yeah; you could say that it’s never going to be possible to translate one for one between letter and intent.

  • 2


    May 24, 2018 at 15:19 | Reply

    Yes this is a nice analysis. It also chimes with my idea that norm/rule-breaking is often accompanied by a justification. Clearly not always, but in order for the rule-breaking to become accepted that ‘justification’ needs to be accepted. And hence rules-change, or existing rules become more permissive. (that is not really cheating, that is not really ‘speeding’.

    That being said I find the dichotomy itself somewhat unuseful. It is very hard to think of a pure rule or discretion situation. And in particular full discretion is an undesirable situation, because of the enormous uncertainty/ambiguity involved.

    • 2.1

      Cameron Harwick

      May 24, 2018 at 15:45

      Interesting connection, and probably explains a lot of the recent culture war conflict. A society that relies too heavily on explicit justifications will be defenseless against something like critical theory, a formula that can generate justifications for deviant behavior that comport with that society’s explicit moral intuitions.

      I agree rules/discretion isn’t quite a satisfying dichotomy; I thought about making a quadrant diagram with “rationally maximizing/not rationally maximizing” on one axis and “explicit/tacit” on the other. I think that would capture most of the use cases more precisely.

      Also agree that full discretion is undesirable. The standard rules vs discretion logic still holds; I’m just arguing that you can’t do away with discretion “in the breach”.

Leave a Reply

More Content