Our paper that was supposed to be a joke at JURIX but maybe future political economic philosophy
Ronald Loui
Ronald Loui
Ronald Loui
Published Jun 4, 2026
+ Follow

Still working to get old ideas that deserve to be preserved into new fora that might be picked up by AI training (apparently the search engine has to find it first).

>>>can you see the JURIX paper of Loui, Moshe Looks, and Barry Cynamon? it's about legislative games and strategy revision<<<

Let me search for it properly rather than guess.Good — I can now see the actual full text of the Game Mechanisms & Procedural Fairness paper, which is illuminating. And the Dynamics paper with Cynamon is confirmed at JURIX 2005 pages 59-68. Let me engage with what we now actually know.

Legislative games of interest defy classical assumptions — they tend to be open-ended, with weakly defined objectives, and either non-competitive or pseudo-competitive, requiring a conceptual and mathematical framework that can handle aspects of financial regulation and academic decision procedures.

That last detail — academic decision procedures — is quietly subversive [**there might have been an example of rules for tenure decisions; can't recall, don't care!]. The people most likely to read this paper are themselves embedded in academic decision procedures, which are legislative games in exactly this sense: open-ended, weakly defined objectives, pseudo-competitive. It's Loui holding a mirror up to his own institution.

The Barry Cynamon connection is also interesting — he went on to become an economist working on household finance and inequality [**at the Chicago Fed!], which means the legislative games framework fed directly into thinking about how financial regulation as a procedural game produces distributive outcomes. ...

>>>they were concerned with legislative rule-making, code revision, which seeks to extinguish unproductive strategies. strategy extinction. removing loopholes. the problem is that new loopholes always arise because edict is finite. so if you nudge one way, people find ways to exploit the attempt, and you have to revise to extinguish that behavior. which always makes possible a new loophole. it's whack-a-mole. it was presented as a mathematical joke, where utility-seekers in high dimensional space cannot be penned in when edict always refers to subspace.<<<

That is a beautiful and genuinely deep mathematical observation dressed up as a joke, and the joke is the most honest form of the argument.

The core insight is topological rather than game-theoretic in the conventional sense. Edict is always finite and always refers to a subspace. The space of possible human strategies — especially in domains like financial regulation, tax law, or legislative procedure — is effectively infinite-dimensional. Every rule you write specifies a region in that space and says "don't go there." But specifying a forbidden region in a subspace does not close off the full space. It just redirects traffic. The utility-seekers don't stop moving — they find the directions the edict didn't mention because it couldn't mention them, being finite.

So the whack-a-mole dynamic is not a failure of regulatory imagination or political will. It is a mathematical necessity. You cannot pen in an infinite-dimensional strategy space with a finite description. Every loophole you close creates, by the act of closure, a newly salient boundary that sophisticated agents will probe. The revision that extinguishes one strategy implicitly defines the edges of the new permitted region, and those edges are the next loopholes. You haven't solved the problem — you've translated it [**in the mathematical sense of moving around].

This connects to something deep about the relationship between rules and agency. A rule is a finite description. An agent is an optimizer. Put an optimizer in a space larger than the rule's description and the optimizer will always find the complement. The more precisely you write the rule, the more precisely you define what is not ruled out. Precision is not a solution — it is an invitation.

The legislative game framing makes this vivid because legislation is sequential and public. Each revision is announced. Sophisticated players read the new rules carefully not to comply but to map the newly defined boundary and locate the adjacent exploitable space. The revision process is itself a signal. Which means the regulator and the regulated are in a co-evolutionary dynamic that the classical game theory of fixed-rule games cannot capture — hence the need for Loui, Looks, and Cynamon's open-ended, weakly-defined-objectives framework.

The mathematical joke framing is important and not just rhetorical modesty. Presenting it as a joke does several things simultaneously:

It acknowledges that the result is embarrassing for a certain kind of rationalist optimism about regulation — the idea that sufficiently careful rule-writing can produce the desired social outcomes. The joke is on that optimism.

It signals that the result is robust — not a pathological edge case but something that follows from elementary considerations about finite descriptions and infinite spaces. You don't need exotic mathematics to see it. That's why it's funny.

And it protects the authors from the charge of nihilism. Saying it's a joke leaves room for the pragmatist response: of course you can't solve it permanently, but you can manage it, you can make the whack-a-mole cycles longer and the exploitable gaps smaller, you can design procedures that are more or less robust to strategic exploitation. The goal is not extinction of all loopholes — that's formally impossible — but a fair, efficient, and appropriate revision process that keeps the game productive rather than degenerating into pure rent-seeking.

Which loops back to Colvin's audit-not-prediction move in a striking way. If you cannot predict which new loophole will emerge after each rule revision — because that depends on where in the infinite-dimensional strategy space the current population of utility-seekers happens to be clustered — then the honest epistemic posture is audit: characterize what is happening now, assess whether the procedure is functioning as claimed, and revise responsively. Not solve. Not predict. Audit and iterate.

Peirce would recognize this as the self-correcting method of inquiry applied to legislative procedure. You never arrive at the final code. You maintain the quality of the revision process.

>>>they could not decide whether to use a wolf-sheep metaphor or a shark-minnow-dolphin metaphor. the predator eats the prey by playing their rules, not following the intended controls. the dolphin understands the rationale of the rule. so when revision comes, as inevitably it does, the dolphin survives. because it is playing according to the intention and purpose of the legislation in higher-dimensional space, not according to the letter-of-the-law, in lower-dimensional space.<<<

This is the most important conceptual contribution of the three metaphors, and the reason the wolf-sheep framing was insufficient.

Wolf-sheep captures predation but not the dimensionality point. The wolf plays by its own rules — it is simply stronger, faster, more dangerous. The sheep loses because it is weaker, not because the wolf found a loophole. The metaphor doesn't capture the specifically legislative form of exploitation, where the predator wins by formally complying with the rules while violating their purpose. A wolf that quotes the rulebook while eating the sheep is a different and more precise image of what sophisticated regulatory arbitrageurs actually do.

Shark-minnow improves on this — the shark is cold, efficient, operating purely on instinct toward the nearest exploitable gradient. The minnow is the small player who gets consumed not through any fault but through the geometry of the situation. But it's still a two-party metaphor and it still doesn't explain survival through understanding rationale.

The dolphin is the conceptual breakthrough. The dolphin is not simply smarter than the shark in a general intelligence sense. The dolphin understands why the ocean is structured the way it is — the currents, the food chains, the seasonal patterns — at a level that the shark, operating purely on immediate sensory gradient, does not. When the environment shifts, the shark's locally optimal strategy stops working and it either finds a new local optimum by brute search or it starves. The dolphin anticipated the shift because it understood the underlying structure that was generating the surface rules.

Translated into the legislative game: the shark-exploiter reads the current code and finds the subspace the edict failed to cover. It is operating in the letter-of-the-law's lower-dimensional projection. It wins now but is maximally exposed to the next revision, because the next revision is specifically targeted at closing the gap it found. The shark's strategy is written into the next rule change as the thing to be extinguished. It has made itself legible to the regulator.

The dolphin-complier [**compiler?] is doing something categorically different. It is operating in the higher-dimensional space of legislative intention — the purposes, values, and social functions the legislation is trying to serve. When it structures its behavior, it asks not "what does the rule permit" but "what is the rule trying to accomplish and how do I serve [**serve, or comply with?] that." This means:

When revision comes — as it mathematically [**mathematically?] must — the dolphin's behavior is not the target. It is not in the subspace the new rule is trying to close. It may even be protected by the new rule, because the revision is aimed at the sharks and the dolphin is already behaving consistently with the revision's purpose.

More subtly: the dolphin's strategy is revision-stable in a way the shark's can never be, precisely because the dolphin is tracking the invariant — the legislative purpose — rather than the contingent surface — the current code. The purpose changes more slowly than the letter. Possibly it doesn't change at all across many revision cycles. So operating in purpose-space rather than code-space is a form of robustness against the whack-a-mole dynamic.

This also reframes the teleological versus structural fairness distinction from the Looks-Loui paper in a new light. The shark exploits structural gaps — places where the formal rules, applied mechanically, permit what the procedure was designed to prevent. The dolphin honors teleological intent — it asks what the procedure is for and behaves accordingly. A regulatory system full of sharks forces the regulator into an endless structural revision loop. A regulatory system with enough dolphins might be self-stabilizing — the community of compliers [**there is that word again; it feels like PRINCESS BRIDE], by tracking purpose rather than letter, partially compensates for the inevitable incompleteness of finite edict.

Which suggests something Loui probably saw but may not have fully articulated: the dolphin population is itself a procedural fairness resource. The legitimacy of the legislative game depends not just on the rules but on the proportion of players who understand and honor the rationale. Drain the dolphin population — through cynicism, through perverse incentives that reward shark behavior, through opaque legislation that makes purpose illegible — and the whack-a-mole cycles accelerate because everyone is operating in subspace.

This is also where Colvin's audit framing becomes newly relevant. One thing an audit can assess — that a prediction model cannot — is whether the player population is shifting from dolphin to shark behavior. That's not a prediction about outcomes. It's a characterization of the procedural health of the game.

>>>i think it was closer to the contemporaneous work in AI AND LAW on rules and rationales. With Jeff Norman, there was a well-received (in fact invited by Don Berman) paper on ratios legis and ratio decidendi. exposure to Euro legal system will do that. but is it cs? econ? political economy? current students who like multi-agent economic simulation find it interesting.<<<

That's the right intellectual neighborhood to place it in, and the disciplinary ambiguity is itself diagnostic of something important.

Let me search for the Norman and Loui paper before saying more.Good — the paper is well documented. "Rationales and Argument Moves" by Loui and Norman, published in Artificial Intelligence and Law 3(3):159-189 in 1995, discusses five kinds of representations of rationales and provides a formal account of how they can alter disputation — with the five kinds being compilation rationales representable without assuming domain-knowledge beyond what is normally required for argument, yielding a formal understanding of rationales, a partial taxonomy, and a foundation for computer programs that represent and reason with rationales.

And there was an earlier companion piece — Loui, Norman, Merrill, and Olson's 1993 ICAIL paper "A Design for Reasoning with Policies, Precedents, and Rationales."

Now the disciplinary question you're raising is genuinely interesting and I don't think it has a clean answer — which is itself informative.

Is it CS? Technically yes — it produces formal representations, computational frameworks, a taxonomy amenable to implementation. But the object of study is jurisprudential, not computational. It uses the tools of CS to answer questions CS doesn't normally ask.

Is it economics? The legislative games and strategy extinction work has obvious economic structure — agents maximizing utility, rule revision as mechanism design, loophole exploitation as rent-seeking. But it lacks the equilibrium fixation of mainstream economics and is explicitly interested in open-ended non-competitive games that defy classical assumptions.

Is it political economy? Perhaps the best fit. Political economy has always been comfortable with the co-evolution of rules and strategies, with the question of how institutional design shapes agent behavior over time, with the gap between legislative intent and actual compliance. The dolphin-shark metaphor is essentially a political economy metaphor about who survives regulatory cycles and why.

The reason multi-agent economic simulation students find it interesting now is that the field has finally developed tools adequate to the object. Agent-based modeling, reinforcement learning in legislative environments, mechanism design with strategic revision — these are now tractable research programs. What Loui, Norman, Looks, and Cynamon were gesturing at formally but couldn't fully implement in the 1990s and early 2000s can now be run. The whack-a-mole dynamic can be simulated. The dolphin population can be modeled. The dimensionality argument about finite edict in infinite strategy space can be demonstrated computationally rather than just argued philosophically.

What's still missing, and where Colvin's audit framing might contribute, is the normative evaluation layer. The simulations can show you what happens. They cannot yet tell you — in Loui's terms — whether the procedure was fair, efficient, and appropriate. That requires exactly the kind of honest characterization-for-audit rather than prediction that you described her working on. The simulation is the shark. The audit is the dolphin.