Traveler's Dilemma¶

Isolate recursion depth as the variable governing whether an iterated-dominance equilibrium predicts behavior — using the same weak-dominance move as the Prisoner's Dilemma chained 98 times, so the $2 prediction evaporates while behavior tracks an incentive gradient the concept is blind to.

Core Idea¶

The Traveler's Dilemma (Kaushik Basu, 1994) is a two-player simultaneous game that demonstrates the behavioral fragility of deep iterated dominance reasoning. Two travelers each had identical antiques damaged by an airline and are asked independently to claim a value between $2 and $100 (the integers). The airline pays both travelers the lower of the two claims; the lower claimant receives a bonus of $2 and the higher claimant pays a penalty of $2. The unique Nash equilibrium, reached by iterated elimination of weakly dominated strategies, is for both travelers to claim $2: any claim above $2 is weakly dominated by a claim one dollar lower, since if the other player claims more, the lower claim wins the bonus, and if the other player claims the same, the lower claim does equally. Chaining this reasoning through all 98 steps from $100 down to $2 yields the equilibrium prediction.

The empirical finding contradicts this prediction sharply. In laboratory experiments by Capra, Goeree, Gomez, and Holt (1999), subjects routinely claimed values near $100 and earned near-maximum payoffs; the Nash equilibrium played almost never. More strikingly, increasing the bonus/penalty parameter shifts behavior systematically: small bonuses produce high claims near $100, large bonuses drive claims toward $2, and the transition is smooth and continuous — behavior responds to the incentive gradient even though the game-theoretic prediction is $2 regardless of the penalty magnitude.

The game is built to make one structural point precise: iterated dominance is a fragile solution concept when the recursion is deep. Each individual elimination step in the chain — claim $99 rather than $100 when you expect the other to claim $100, claim $98 rather than $99 when you expect them to claim $99, and so on — is locally plausible, but the chain requires that both players perform 98 rounds of correct forward reasoning about each other's reasoning, and that both believe the other will do so as well. In practice, agents do not perform dominance elimination at depth; they act at shallow reasoning depths, and the Quantal Response Equilibrium framework (McKelvey and Palfrey, 1995, 1998) accounts for the observed behavior by injecting small choice errors that unravel the deep chain, predicting the observed high-claim concentration and its sensitivity to the bonus parameter.

Structural Signature¶

Sig role-phrases:

the two simultaneous claimants — players who independently name a value on a fine-grained bounded grid ($2–$100)
the undercut-rewarding rule — both paid the lower claim, with a small bonus to the lower claimant and penalty to the higher, so claiming one dollar lower weakly dominates
the deep iterated-dominance chain — the unique Nash equilibrium ($2, $2) reached only by chaining the identical weak-dominance step 98 times, each player reasoning about the other's reasoning at full depth
the behavioral evaporation — the demonstration's payload: subjects claim near the upper bound and the Nash prediction is played almost never, so the equilibrium is logically derivable but not behaviorally predictive
the recursion-depth parameter — the governing variable isolated by holding the logical move fixed against the one-step Prisoner's Dilemma: shallow chains predict well, deep chains evaporate
the parameter-blindness dissociation — the bonus/penalty magnitude leaves the $2 prediction unchanged yet slides observed behavior smoothly from the upper bound toward the floor, exposing a gradient the standard concept cannot see
the error-tolerant repair — the completion: Quantal Response Equilibrium / level-k lets small choice errors unravel the deep chain, recovering both the high-claim concentration and its sensitivity to the penalty

What It Is Not¶

Not a one-step dominance argument. Its defining feature is not that a dominated strategy exists but the depth of the chain: the same weak-dominance move iterated 98 times from $100 down to $2. A Prisoner's-Dilemma-style single elimination predicts behavior well; what the Traveler's Dilemma isolates is precisely the recursion depth that the one-shot case lacks.
Not a refutation of equilibrium analysis as such. It does not show that Nash reasoning is worthless; it shows that equilibria sitting behind deep inferential chains are behaviorally fragile. Shallow-chain equilibria remain sound forecasts, and the repair is an error-tolerant equilibrium concept, not the abandonment of equilibrium analysis.
Not a demonstration that subjects are irrational. Players claiming near $100 earn near-maximum payoffs; it is the iterated-dominance prediction of $2 that fails, not the players. They reason at shallow depth and do well by it, so "irrationality" mislabels what is better read as the equilibrium over-demanding recursion no agent traverses.
Not a game whose outcome is insensitive to incentives. The bonus/penalty magnitude moves observed behavior smoothly from the upper bound toward the floor. What is parameter-blind is the Nash prediction (always $2); the dissociation between an inert prediction and a responsive behavior is the whole point, not a claim that incentives do not matter.
Not a flawed or mis-derived equilibrium. The ($2, $2) outcome is the correct unique Nash equilibrium under iterated weak dominance — the logic is valid. The lesson is the gap between an equilibrium being logically derivable and being behaviorally predictive, not an error in the derivation.

Scope of Application¶

The Traveler's Dilemma lives within one substrate — strategic games solved by iterated dominance, populated by agents reasoning (or failing to reason) about each other — and specifically the settings sharing its load-bearing feature, a deep chain of iterated weak-dominance eliminations driven by a small incentive to undercut. Its reach is unusually narrow even for a domain-specific abstraction: as a purpose-built counterexample its value is largely internal to game theory, and the general fragility-of-deep-recursion lesson travels via nash_equilibrium / bounded_rationality, not by importing the airline-and-antiques construction.

Game-theory pedagogy and equilibrium refinement — the home use, the canonical exhibit (beside the centipede game) that iterated dominance is fragile when recursion depth is large.
Behavioral economics — an empirical test bed for bounded rationality, level-k thinking, and Quantal Response Equilibrium (McKelvey-Palfrey), which recovers the observed high-claim behavior by injecting small choice errors that unravel the deep chain.
Industrial organization (Bertrand competition) — the same structure in another dress, where each firm slightly undercuts and the chain collapses to marginal-cost pricing that real markets fall short of in exactly the anticipated way.
Auction and pricing design — any mechanism rewarding undercutting by a small margin on a finite grid builds the identical deep chain, so its iterated-dominance equilibrium is behaviorally fragile.
The centipede game (sequential cousin) — the same depth-driven critique carried in a sequential tree rather than a simultaneous move.

Clarity¶

Naming the Traveler's Dilemma makes legible a property of iterated dominance that single-step examples hide: that the solution concept's behavioral validity degrades with the depth of recursion it demands. The Prisoner's Dilemma settles in one round of elimination and predicts behavior well; the Traveler's Dilemma uses the identical logical move but chains it 98 times, and the prediction evaporates. Holding the two side by side dissolves the assumption that iterated dominance is uniformly trustworthy — it shows that number of elimination steps is itself a parameter governing whether the equilibrium is a sound forecast or an artifact of taking the recursion to a depth no agent actually traverses. The sharper question it lets a game theorist ask is therefore not merely "what is the Nash equilibrium?" but "how deep is the dominance chain that produces it, and is that depth behaviorally plausible?"

The distinction it sharpens is between an equilibrium being logically derivable and being behaviorally predictive — between what common knowledge of rationality entails in principle and what boundedly rational agents actually do. The game makes this gap measurable rather than rhetorical: the bonus/penalty parameter has no effect whatsoever on the iterated-dominance prediction, which is $2 regardless, yet it moves observed behavior smoothly from the upper bound toward the floor. That dissociation localizes the failure precisely — the standard concept is blind to an incentive gradient that demonstrably drives choices — and points to the right repair: a model like Quantal Response Equilibrium that lets small choice errors unravel deep chains and recovers the gradient sensitivity. Naming the dilemma thus equips a practitioner to distrust deep-recursion equilibria specifically, rather than abandoning equilibrium analysis wholesale.

Manages Complexity¶

Whether an iterated-dominance equilibrium is a trustworthy forecast had been, in practice, a case-by-case verdict: the Prisoner's Dilemma's one-round elimination predicts behavior well, Bertrand competition's undercutting chain collapses to marginal-cost pricing that markets often do not reach, the centipede game's backward unraveling fails empirically — each game judged on its own track record, with no parameter saying in advance which way it would go. The Traveler's Dilemma compresses that scattered judgment onto a single scalar: the depth of the dominance chain the equilibrium requires. By using the identical logical move as the Prisoner's Dilemma but chaining it 98 times, it isolates recursion depth as the variable that governs behavioral validity, holding everything else fixed. The analyst no longer asks the open-ended "will agents actually play this equilibrium?" but reads the answer off one number — how many elimination steps separate the action space from the predicted outcome — with shallow chains behaviorally sound and deep chains evaporating into artifacts of a recursion no agent traverses. A diffuse "iterated dominance is sometimes unreliable" becomes a measurable threshold on a single tracked quantity.

That scalar sits inside a clean two-axis decomposition that tells the analyst not just whether the prediction fails but which way behavior moves and how to model it. The game separates two things the standard concept fuses — an equilibrium being logically derivable and being behaviorally predictive — and makes the gap quantitative by exhibiting a parameter, the bonus/penalty magnitude, that the iterated-dominance prediction is completely blind to (it is $2 at every penalty) yet that drives observed claims smoothly from the upper bound toward the floor. So the practitioner tracks two quantities: recursion depth, which sets whether the equilibrium is a forecast or an artifact, and the incentive gradient, which sets where boundedly-rational behavior actually lands. From those, the qualitative outcome and its remedy both follow without re-deriving each case: deep chain plus shallow penalty → behavior near the upper bound, modeled by letting small choice errors (Quantal Response Equilibrium) unravel the chain and recover the gradient; steepening the penalty slides behavior toward the floor. The broad, hand-waved critique of "rationality is unreasonable at depth" collapses to a depth parameter, a gradient parameter, and a known repair keyed to them.

Abstract Reasoning¶

Within game theory and equilibrium refinement the game licenses reasoning moves that all run on the depth of the iterated-dominance chain and the gap between logical derivability and behavioral prediction.

Diagnostic — count the recursion depth that produces an equilibrium, and read its behavioral validity off that depth. The signature move treats the number of elimination steps as itself a parameter: the analyst reasons FROM "this equilibrium is reached only by chaining the same weak-dominance move many times — claim one dollar lower than whatever the other claims, 98 rounds from $100 down to $2" TO "the prediction requires both players to perform that many rounds of correct mutual reasoning, and to believe the other does too." A second diagnostic move converts depth into a forecast about the forecast: reasoning FROM "the dominance chain is deep" TO "the equilibrium is likely a behavioral artifact of a recursion no agent actually traverses," and FROM "the chain is one step (as in the Prisoner's Dilemma)" TO "the equilibrium is a sound forecast." A third diagnostic move detects the tell-tale dissociation: reasoning FROM "an incentive parameter (the bonus/penalty magnitude) moves observed behavior yet leaves the iterated-dominance prediction unchanged" TO "the standard concept is blind to a gradient that demonstrably drives choices." The move is FROM the depth and parameter-blindness of the equilibrium TO a verdict on whether it will predict behavior.

Interventionist — model deep-recursion games with error-tolerant equilibria, and steepen the penalty to move where behavior lands. The corrective move repairs the prediction rather than discarding equilibrium analysis: reasoning FROM "deep chains unravel under small choice errors" TO "use Quantal Response Equilibrium (or level-k reasoning), which injects errors that break the recursion and recovers both the observed high-claim concentration and its sensitivity to the bonus parameter." The analyst reasons FROM "this equilibrium rests on deep iterated dominance" TO "model it with an error-tolerant solution concept, not pure best-response." A second interventionist move acts on the incentive gradient the standard concept ignores: reasoning FROM "increase the bonus/penalty magnitude" TO "boundedly-rational claims slide smoothly from near the upper bound toward the floor," so a mechanism designer who wants behavior near the cooperative outcome keeps the penalty shallow, and one who wants the competitive floor steepens it — predicting the behavioral effect of a parameter to which the Nash prediction is wholly insensitive.

Boundary-drawing — separate logically-derivable from behaviorally-predictive equilibria, and make recursion depth the governing variable. A first boundary move splits two things the standard concept fuses: an equilibrium being logically derivable (what common knowledge of rationality entails in principle) versus behaviorally predictive (what boundedly rational agents actually do). The analyst reasons FROM "is this equilibrium derivable, or is it observed?" TO "deep-recursion equilibria can be the first without the second," and the game makes the gap measurable because the penalty parameter dissociates them — no effect on the $2 prediction, large effect on behavior. A second boundary move isolates recursion depth as the controlling variable by holding the logical move fixed: the Prisoner's Dilemma settles in one elimination round and predicts well, the Traveler's Dilemma uses the identical move chained 98 times and the prediction evaporates, so the analyst reasons FROM "how many elimination steps separate the action space from the predicted outcome?" TO "whether the equilibrium is trustworthy" — and distinguishes the dilemma from its sequential cousin the centipede game (same critique, a tree rather than a simultaneous move) and from race-to-the-bottom dynamics with different mechanisms.

Predictive — a deep chain with a shallow penalty forecasts behavior near the upper bound, and the gradient forecasts the direction of departure. A forward move predicts the experimental result the equilibrium misses: reasoning FROM "the dominance chain is deep and the penalty small" TO "subjects will claim values near $100 and earn near-maximum payoffs, with the Nash equilibrium played almost never." A second predictive move forecasts the response to a parameter the standard concept treats as inert: reasoning FROM "the iterated-dominance prediction is $2 at every penalty, but each elimination step's pull strengthens as the penalty grows" TO "behavior will move smoothly from the upper bound toward the floor as the bonus/penalty rises," predicting a continuous transition where the equilibrium predicts a constant. A third predictive move generalizes the fragility: reasoning FROM "any mechanism that rewards undercutting by a small margin builds the same deep chain" TO "its iterated-dominance equilibrium (Bertrand marginal-cost pricing, the race to the bottom on finite grids) will be behaviorally fragile in the same way," forecasting that real markets and agents will fall short of the deeply-recursive prediction.

Knowledge Transfer¶

Within game theory the Traveler's Dilemma transfers as mechanism to the settings that share its load-bearing feature — a deep chain of iterated weak-dominance eliminations driven by a small incentive to undercut. The diagnostic (count the recursion depth; read behavioral validity off it), the dissociation it exhibits (an incentive parameter that moves behavior while leaving the iterated-dominance prediction fixed), and the repair (model with an error-tolerant equilibrium such as Quantal Response or level-k, which lets small choice errors unravel the chain and recover the gradient) carry to its near neighbors. Bertrand price competition is the same structure in industrial-organization dress: each firm slightly undercuts, the chain collapses to marginal-cost pricing, and real markets fall short of that deeply-recursive prediction in exactly the way the Traveler's Dilemma anticipates. Any auction or pricing mechanism that rewards undercutting by a small margin builds the same chain on a finite grid, and the centipede game is the sequential-tree cousin carrying the identical critique. Across these the transfer is genuine recognition of the same depth-driven fragility, not analogy — but they are all members of one substrate: strategic games solved by iterated dominance, populated by agents reasoning (or failing to reason) about each other.

Beyond game theory and its adjacent fields, the honest report is that the construction itself does not travel at all. There is no biological, physical, or computational "Traveler's Dilemma" — no substrate outside strategic interaction has two agents naming numbers between bounds under a bonus/penalty rule — so unlike many domain-specific abstractions there is not even a productive analogy to mark; the object is a purpose-built counterexample, and its content is exhausted by the lesson it was constructed to deliver. This is worth stating plainly because it is the unusual case: the entry's value is almost entirely internal to the field, as a demonstration that makes a general critique vivid and measurable, rather than as a pattern recognized across domains.

What does carry the lesson is not the named game but the general patterns it instantiates, and the cross-domain reach belongs entirely to them. The fragility it dramatizes — that an equilibrium can be logically derivable yet not behaviorally predictive, and that the gap widens with the depth of recursion the equilibrium demands — is a property of the parent solution concept and is carried by nash_equilibrium together with bounded_rationality and the broader equilibrium-refinement discussion; the deviation it documents is the general fact that real agents reason at shallow depth, which bounded_rationality already houses. When the lesson is needed elsewhere — designing a mechanism whose intended equilibrium sits behind a long inferential chain, or distrusting a prediction that requires many rounds of mutual reasoning — it should be carried by "deep iterated reasoning is behaviorally fragile, model it with bounded-rational equilibria," not by importing "Traveler's Dilemma," whose airline, antiques, and $2–$100 grid are pure scaffolding for the demonstration. The general fragility-of-deep-recursion pattern travels via the parent primes; the named construction stays home as the example that proves it — the division Structural Core vs. Domain Accent makes precise below.

Examples¶

Canonical¶

Kaushik Basu's 1994 construction shows the chain concretely. Two travelers each claim an integer from $2 to $100; the airline pays both the lower claim, adds a $2 bonus to the lower claimant, and docks $2 from the higher. Consider claiming $100. If your opponent also claims $100 you receive $100 — but if you claim $99 instead, then whenever the opponent claims $100 the lower figure is $99, and as the low claimant you receive $99 + $2 = $101 (the opponent gets $99 − $2 = $97); if the opponent also claims $99 you tie at $99. So $99 weakly dominates $100. The identical logic makes $98 dominate $99, and so on down 98 steps to the unique Nash equilibrium of ($2, $2). Yet in the laboratory experiments of Capra, Goeree, Gomez, and Holt (1999), subjects claimed near $100 and earned near-maximum payoffs; the $2 equilibrium was played almost never.

Mapped back: The two travelers naming integers on the $2–$100 grid are the two simultaneous claimants, and the pay-the-lower-plus-bonus/penalty structure is the undercut-rewarding rule that makes each dollar-lower claim weakly dominant. Chaining that move 98 times to ($2, $2) is the deep iterated-dominance chain, and subjects claiming near $100 instead is the behavioral evaporation.

Applied / In Practice¶

Bertrand price competition is the same fragility playing out in real markets. Two firms selling an identical product each have an incentive to undercut the other by a penny to capture the whole market, and iterating that logic — the textbook Bertrand argument — drives price all the way down to marginal cost, leaving both firms with zero profit even though there are only two of them. This is the "Bertrand paradox," and real duopolies routinely fail to reach it: firms selling near-identical goods commonly sustain prices well above marginal cost. The deeply recursive undercutting equilibrium is logically derivable but behaviorally unreached, exactly as the Traveler's Dilemma predicts, and economists model the shortfall with the same error-tolerant and bounded-rationality tools, or with repeated-game and capacity refinements.

Mapped back: Each firm's incentive to shave the price is the undercut-rewarding rule, and iterating undercuts to marginal cost is the deep iterated-dominance chain. Real prices holding above marginal cost is the behavioral evaporation — the equilibrium reached only through a recursion depth actual firms do not traverse, the recursion-depth parameter governing whether the prediction binds.

Structural Tensions¶

T1: Correct equilibrium versus useless forecast (the derivation is valid and the prediction still fails). The ($2, $2) outcome is the genuinely correct unique Nash equilibrium under iterated weak dominance — the logic is impeccable, no step is mis-derived. And it is a bad forecast: subjects claim near $100 and the equilibrium is played almost never. The practitioner must therefore hold two things that feel contradictory — the analysis is right, and the analysis is not to be trusted here — without collapsing to either "the math is wrong" (it is not) or "equilibrium analysis is worthless" (shallow-chain equilibria predict well). The concept lives exactly on the fault line where mathematical validity and empirical predictive power come apart, and its lesson is precisely that a correctly-derived equilibrium can be the wrong thing to believe about behavior. Diagnostic: Is the equilibrium being distrusted because the derivation is flawed (it is not), or because a valid derivation sits behind a recursion too deep to be behaviorally predictive?

T2: Depth as the governing scalar versus the incentive gradient it cannot see (one variable is not enough). The game's headline achievement is isolating recursion depth as the variable that decides behavioral validity, by holding the logical move fixed against the one-step Prisoner's Dilemma. But its own most striking finding is that a second variable — the bonus/penalty magnitude, to which the Nash prediction is wholly blind — moves behavior smoothly from the upper bound toward the floor. So depth alone does not fix where boundedly-rational play lands: a deep chain with a shallow penalty leaves behavior near $100, while steepening the penalty drives it toward $2 at the same depth. The concept that teaches "count the depth" simultaneously demonstrates that depth is insufficient, because the per-step incentive strength co-determines the outcome. The scalar the game isolates and the gradient it exposes are both load-bearing, and neither predicts alone. Diagnostic: Is the behavioral forecast resting on recursion depth alone, or is the per-step incentive gradient (penalty magnitude) — which Nash ignores — being tracked as an independent determinant of where behavior lands?

T3: The over-demanding solution concept versus the exploitability of shallow reasoning (who is really irrational). The game's charitable inversion is that subjects claiming near $100 are not irrational — they earn near-maximum payoffs, and it is the equilibrium that over-demands recursion no agent traverses. That correction is right here, but it does not license the general claim that shallow reasoning is always fine: in settings where a counterpart does reason deeply, the shallow agent who ignores the dominance chain can be systematically exploited, and the very charity that rescues the travelers would mislead an agent facing a sophisticated adversary. So the game establishes that the solution concept can over-demand depth and leaves open that under-supplying depth is dangerous elsewhere, and reading its lesson as "deep reasoning is unnecessary" overshoots. Whether shallow play is wisdom or vulnerability depends on the counterpart, which the single game does not settle. Diagnostic: Does shallow reasoning do well here because the counterpart also reasons shallowly, or would a deeper-reasoning opponent exploit the shallow player — making the depth the equilibrium demands actually necessary?

T4: The error-tolerant repair versus its descriptive flexibility (fitting behavior can cost predictive discipline). Quantal Response Equilibrium (and level-k) repair the prediction by injecting small choice errors that unravel the deep chain, recovering both the high-claim concentration and its sensitivity to the penalty — a genuine advance over a Nash prediction that is simply wrong. But the repair buys its fit with free parameters (an error or rationality-precision term), and a model tuned by a free parameter risks being flexible enough to accommodate a wide range of outcomes, trading the crisp, falsifiable (if wrong) Nash point prediction for a curve that fits after the fact. The move from a sharp incorrect forecast to a tunable accurate one gains realism and can lose the predictive discipline of a parameter-free theory. Diagnostic: Is the error-tolerant model predicting behavior out of sample, or recovering the observed data by fitting a free error parameter that could have accommodated a different result too?

T5: Autonomy versus reduction (a purpose-built counterexample or the parent lesson it dramatizes). More sharply than most entries, the Traveler's Dilemma does not travel even by analogy: no substrate outside strategic interaction has two agents naming numbers between bounds under a bonus/penalty rule, and the airline, antiques, and $2–$100 grid are pure scaffolding for a demonstration whose content is exhausted by the lesson it delivers. What carries is the parent — that an equilibrium can be nash_equilibrium-derivable yet not behaviorally predictive, and that bounded_rationality means real agents reason at shallow depth, so deep-recursion equilibria are fragile. When the lesson is needed (a mechanism whose intended equilibrium sits behind a long inferential chain), it should be carried as "deep iterated reasoning is behaviorally fragile; model it with bounded-rational equilibria," not by importing the named game. The tension is unusually one-sided: the construction's vividness and its non-portability are the same artificiality, so its value is almost entirely internal to game theory. Diagnostic: Resolve toward the parents (nash-equilibrium fragility plus bounded rationality) whenever the lesson is wanted outside the game; toward the Traveler's Dilemma only as the constructed exhibit that makes the depth-fragility point vivid and measurable.

Structural–Framed Character¶

The Traveler's Dilemma sits at the framed pole of the spectrum, and unusually so even among framed entries — it is a purpose-built artifact of a theoretical tradition whose content is exhausted by the lesson it was constructed to deliver. The five criteria run overwhelmingly framed. On human_practice_bound it is maximal in the constitutive sense: the object is a designed scenario — two agents naming numbers on a bounded grid under a bonus/penalty rule — that has no existence apart from the practice of strategic-reasoning analysis; strip away the game-theoretic setting and there is nothing left, not even a mechanism running observer-free the way a fault slips or a market clears, because the airline, the antiques, and the $2–$100 grid are scaffolding for a demonstration, not features of any substrate the world supplies. Its institutional_origin is total and datable: Basu constructed it in 1994 as a counterexample, the QRE/level-k repair (McKelvey–Palfrey) is apparatus from the same tradition, and the whole object is furniture of game theory and behavioral economics, invented rather than discovered. On vocab_travels it fails completely — the operative content does not float free of strategic interaction, and the entry itself is candid that the construction "does not travel even by analogy," with no biological, physical, or computational counterpart. On import_vs_recognize it patterns as within-domain recognition only: Bertrand competition and the centipede game are recognized as the same depth-driven fragility, but beyond that adjacent family there is nothing to import into. Evaluative_weight is the one criterion that is not maximal — the game does not convict a move the way "ad hominem" does; its payload is a methodological caution about when to trust a solution concept rather than a verdict on an agent (indeed its charitable reading is that the near-$100 players are not irrational). But that caution is still normatively loaded (it tells the analyst which equilibria to distrust), so even here it leans framed rather than toward neutral mechanism-description. Net placement: framed pole, effectively a constructed teaching exhibit rather than a mechanism read off nature.

The portable structural skeleton is singular and, tellingly, not proprietary to the game at all: a fixed point that is logically derivable can fail to be behaviorally predictive, and the gap widens with the depth of iterated mutual reasoning the fixed point demands. That skeleton is exactly what the Traveler's Dilemma instantiates from its parent primes — nash_equilibrium (the derivable fixed point) together with bounded_rationality (real agents reason at shallow depth) — and the entry locates the cross-domain reach entirely in those parents: when the lesson is wanted elsewhere it should be carried as "deep iterated reasoning is behaviorally fragile; model it with bounded-rational equilibria," never by importing the named construction. The named game's distinctive content — the 98-step chain, the parameter-blindness dissociation, the airline-and-antiques dressing — is precisely the part that stays home and does not travel. Its character: a constructed, practice-constituted counterexample whose vividness and its total non-portability are the same artificiality, structural only in the derivable-fixed-point-versus-bounded-behavior skeleton it borrows from nash_equilibrium and bounded_rationality and stages as a methodological verdict.

Structural Core vs. Domain Accent¶

This section decides why the Traveler's Dilemma is a domain-specific abstraction and not a prime — and it is the rare entry where the case is nearly overdetermined, because the construction does not even travel by analogy, only its parent lesson does.

What is skeletal (could lift toward a cross-domain prime). Strip the airline and antiques and a thin abstract structure survives: a fixed point that is logically derivable from mutual best-response can fail to be behaviorally predictive, and the gap between the two widens with the depth of iterated mutual reasoning the fixed point demands. Stated abstractly, its portable pieces are two — a derivable equilibrium (a self-consistent profile every agent would confirm given the others), carried by nash_equilibrium, and the fact that real agents reason to shallow depth rather than to logical closure, carried by bounded_rationality. From just those two the whole lesson follows: an equilibrium reached only by chaining a mutual-reasoning step many times sits behind a recursion no agent traverses, so its predictive authority decays with chain length. That skeleton is genuinely substrate-portable — it governs any mechanism whose intended equilibrium rests on a long inferential chain — which is precisely why it recurs as the two parent primes; but it is the shared core, not what makes the Traveler's Dilemma distinctive.

What is domain-bound. Everything that makes the object the Traveler's Dilemma in particular is game-theoretic scaffolding purpose-built for a demonstration, and none of it survives extraction. The two simultaneous claimants naming integers on a $2–$100 grid; the pay-the-lower-plus-$2-bonus/minus-$2-penalty rule engineered so that claiming one dollar lower weakly dominates; the specific 98-step chain from $100 to $2; the parameter-blindness dissociation (the penalty magnitude leaves the Nash prediction inert at $2 while sliding observed behavior smoothly toward the floor); and the Capra–Goeree–Gomez–Holt experimental protocol and the QRE/level-k repair apparatus (McKelvey–Palfrey) that recovers the gradient — all of it is furniture of game theory and behavioral economics, invented rather than discovered. The decisive test is unusually stark: remove the strategic-reasoning setting and there is nothing left at all — not even a looser thing — because no substrate in nature has two agents naming numbers between bounds under a bonus/penalty rule. The construction is exhausted by the lesson it was built to make vivid.

Why this does not clear the prime bar. A prime's vocabulary travels and its cross-domain transfer is recognition of the same mechanism, not analogy. The Traveler's Dilemma's transfer is bimodal in an especially one-sided way. Within game theory and its adjacent fields the mechanism travels intact by genuine recognition: Bertrand price competition is the same depth-driven undercutting chain in industrial-organization dress (real duopolies hold prices above marginal cost exactly as the dilemma anticipates), any auction rewarding small-margin undercutting on a finite grid builds the identical chain, and the centipede game carries the same critique in a sequential tree — all members of the one substrate of strategic games solved by iterated dominance. Beyond that family the named construction does not travel even by analogy — there is no biological, physical, or computational "Traveler's Dilemma" to recognize or to borrow — so its entire cross-domain value is internal, as the exhibit that makes a general critique measurable. When the bare lesson is needed elsewhere — designing a mechanism whose intended equilibrium sits behind a long inferential chain, or distrusting a forecast that presumes many rounds of mutual reasoning — it is already carried, in fully general form, by nash_equilibrium (the derivable fixed point) together with bounded_rationality (shallow-depth reasoning), as "deep iterated reasoning is behaviorally fragile; model it with bounded-rational equilibria." The cross-domain reach belongs entirely to those parents; the named game's distinctive content — the 98-step chain, the dissociation, the airline-and-antiques dressing — is exactly the part that stays home. The Traveler's Dilemma clears the domain-specific bar as a canonical construction of game theory, but it carries no substrate-spanning content of its own that the parent primes do not already carry more generally.

Relationships to Other Abstractions¶

Current abstraction Traveler's Dilemma Domain-specific

Parents (2) — more general patterns this builds on

Traveler's Dilemma is part of Dominated Strategy Domain-specific

Traveler's Dilemma contains a 98-step chain of weakly dominated claims whose iterative elimination produces its behaviorally fragile endpoint.
Traveler's Dilemma is a decomposition of Bounded Rationality Prime

Removing the claim game leaves the limited-depth principle that a valid inference chain can cease to predict behavior when it exceeds agents' reasoning budget.

Not to Be Confused With¶

Prisoner's Dilemma. The canonical two-player game whose dominant-strategy equilibrium is reached in a single round of dominance elimination — and the deliberate foil the Traveler's Dilemma is built against. Both use the identical weak/strict-dominance move, but the Prisoner's Dilemma settles at depth one and predicts behavior well, whereas the Traveler's Dilemma chains that same move 98 times and the prediction evaporates. The entire demonstration is holding the two side by side to isolate recursion depth as the governing variable. Tell: does one elimination step reach the equilibrium (Prisoner's Dilemma, behaviorally sound), or does it take a long chain of mutual reasoning to get there (Traveler's Dilemma, behaviorally fragile)?
Centipede game. The sequential-tree cousin carrying the identical depth-driven critique — backward induction unravels cooperation over many stages, and real players deviate from the induced equilibrium. It shares the Traveler's Dilemma's lesson (deep iterated reasoning is behaviorally fragile) but stages it in a sequential game with alternating moves rather than a simultaneous one-shot claim. Tell: do players move in turn down a tree with backward induction supplying the depth (centipede), or name numbers simultaneously with iterated dominance supplying it (Traveler's Dilemma)?
Bertrand competition / Bertrand paradox. The industrial-organization setting where two firms undercut each other's price until it collapses to marginal cost — the same deep undercutting chain in market dress, and real duopolies fall short of the prediction exactly as the Traveler's Dilemma anticipates. It is a within-domain sibling recognized as the same mechanism, not a distinct concept, but keyed to prices and firms rather than antique claims. Tell: is the undercutting chain about competing firms' prices reaching marginal cost (Bertrand), or about two claimants naming values under a bonus/penalty rule (Traveler's Dilemma)? — same fragility, different substrate within economics.
Race to the bottom. A superficially similar downward spiral — jurisdictions, firms, or agents competitively lowering some standard (taxes, wages, regulation) toward a mutually harmful floor. It shares the picture of iterated undercutting driving outcomes down, but its mechanism is competitive pressure among many actors, not the deep iterated weak-dominance recursion whose behavioral fragility is the Traveler's Dilemma's whole point. Tell: is the downward movement driven by many actors responding to competitive pressure (race to the bottom), or by a formal chain of mutual dominance reasoning that real agents fail to traverse (Traveler's Dilemma)?
Quantal Response Equilibrium / level-k reasoning. The repair, not the game — error-tolerant solution concepts that inject small choice errors (QRE) or finite reasoning depth (level-k) to unravel the deep chain and recover the observed high-claim behavior and its penalty-sensitivity. These are the modeling apparatus applied to the Traveler's Dilemma, not the dilemma itself; confusing them mistakes the diagnostic instrument for the phenomenon it diagnoses. Tell: is it the constructed scenario exhibiting the prediction failure (the dilemma), or the bounded-rational equilibrium concept brought in to model the failure (QRE/level-k)?
The nash_equilibrium + bounded_rationality umbrella (parent primes). The substrate-neutral lesson the game instantiates — a logically derivable fixed point can fail to be behaviorally predictive, and the gap widens with the depth of iterated mutual reasoning it demands. This is the portable content, carried by nash_equilibrium (the derivable fixed point) and bounded_rationality (agents reason at shallow depth); it is what actually travels when the lesson is needed elsewhere, whereas the named game is pure scaffolding. Tell: are you carrying the general point "deep iterated reasoning is behaviorally fragile, model it with bounded-rational equilibria" (the parents), or the specific airline-antiques-98-step construction (the dilemma, which does not travel even by analogy)? (Treated fully in a later section.)

Neighborhood in Abstraction Space¶

Traveler's Dilemma sits in a crowded region of the domain-specific corpus (0^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Strategic Interaction & Game Theory (23 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12