Mathematical game theory

Introduction

Mathematical Game Theory: A Proper Axiomatization

Mathematical game theory—as formalized by John von Neumann, Oskar Morgenstern, and John Nash—defines rationality in terms of computable optimization. This is not a stylistic preference but a necessary condition for real-world applicability: an optimal strategy must be computable. This requirement is both a logical and empirical constraint. For game theory to inform the actions of any rational agent, a strategy's optimality must be provable—that is, its correctness must be derivable through computation.

Any assertion whose truth value cannot be evaluated by a deductive system—specifically, classical first-order logic (CFOL)—is, by necessity, inaccessible to rational verification and decision-making. While such assertions may appear meaningful in informal discourse or statistical estimation, their practical validity depends on foundational verifiability within a formal system. For any inference—such as Bayes' rule—to be useful to rational agents, its underlying structure must be reducible to provable rules within CFOL.

Unprovable claims may be false, undecidable, or uncomputable, and thus fall outside the scope not only of applied economics, but of any formal theory of rational action. In such contexts, only verifiable knowledge has strategic utility.

In mathematics, a proof is a finite, mechanically checkable sequence of statements that conforms to well-defined rules of inference. Whether verified by a human or executed by a machine, a proof is computation—in every case, without exception.

Therefore:

• All valid proofs must be verifiable for correctness; they must be checkable for truth via computation by any number of rational agents.

• By the Church–Turing Thesis, all effective procedures are equivalent in computational power to a Turing machine
  (Church, A. “An Unsolvable Problem of Elementary Number Theory,” American Journal of Mathematics, 58(2), 1936).

Consequently, anything beyond this boundary is not merely unproven—it is unverifiable, and hence epistemically null.

This is not a limitation. It is a foundational filter that separates meaningful from non-meaningful claims in applied mathematical theory. If a proposition cannot be verified computationally, then it cannot inform rational strategy, cannot be integrated into a formal model, and cannot be acted upon. It is discarded—not by subjective judgment, but by necessity.

This principle applies not only to bounded agents, but to any conceivable agent—human, artificial, or hypothetical—regardless of memory, intelligence, or computational capacity.

Even a hypothetical epistemic super-agent cannot rationally act on unprovable claims, because usefulness presupposes verifiability, and verifiability requires computation.

Why Classical First-Order Logic (CFOL)?

While theorists may employ higher-order, modal, or probabilistic frameworks—such as set theory, dynamic epistemic logic, or Bayesian inference—every tractable form of reasoning reducible to a computable process can, in principle, be expressed within classical first-order logic (CFOL). This is due to CFOL’s status as the minimal complete deductive system capable of encoding any Turing-computable function, as demonstrated by the foundational equivalence results of Gödel, Church, and Turing. Though alternative frameworks may offer expressive convenience or semantic clarity in specific domains, they do not extend beyond the computational boundaries that CFOL can formally represent.

The Foundational Status of CFOL Rests on Three Pillars:

1. Soundness and Completeness (Gödel, 1930)

• Soundness: Every provable CFOL statement is logically valid (i.e., true in all models).

• Completeness: Every universally valid CFOL statement is provable within the system.

• CFOL proofs are finite, syntactic, and mechanically verifiable—rendering them executable by both human reasoners and formal machines.

2. Computational Universality (Church–Turing Thesis)

• Turing machines, the lambda calculus, and CFOL-definable functions are computationally equivalent in expressive power.

• Any meaningful result in game theory—such as equilibrium existence, best response derivation, or Bayesian inference—must be computable to qualify as rational.

• Therefore, rational strategy presupposes Turing-computable procedures, all of which CFOL can formally encode.

3. Computability Encoded in CFOL

• Although CFOL does not natively express higher-order constructs such as common knowledge or belief hierarchies, it can encode any computable function via techniques such as Gödel numbering and arithmetization of syntax.

• This renders CFOL sufficient for modeling any behavior with decision-theoretic relevance that is computationally realizable by an agent.

Defining Rationality in Formal Terms

Within this framework, a rational agent is not an idealized philosophical abstraction, but a formally specified system. Precisely, a rational agent is:

A system that:

• Operates under bivalence (each proposition is either true or false; no third option is allowed under the law of excluded middle and the law of non-contradiction),

• Applies inference deterministically from explicit rules,

• Selects strategies via provable, computable optimization.

In short:

A rational agent is a Turing machine with a utility function.

This is not a metaphor. It is a formal specification of what it means to act rationally.

Von Neumann’s Actual Insight

Von Neumann did not merely invent game theory—he recognized its computational essence. He understood that strategy, optimization, and equilibrium are not metaphysical abstractions, but finite algorithms to be executed by real agents, under real constraints, in real time.

He did not merely formalize games in logic—
He built the first computer to run them.

Recommended References

• Gödel, K. (1930). The completeness of the axioms of the functional calculus of logic.

• Church, A. (1936). An unsolvable problem of elementary number theory. American Journal of Mathematics, 58(2).

• Turing, A. M. (1936). On computable numbers, with an application to the Entscheidungsproblem. Proceedings of the London Mathematical Society, 2(42).

• von Neumann, J., & Morgenstern, O. (1944). Theory of Games and Economic Behavior.

• Nash, J. (1950). Equilibrium points in n-person games. Proceedings of the National Academy of Sciences, 36(1).

Rationality, Strategic Uncertainty, and Logical Determinism

In canonical examples such as the Prisoner’s Dilemma, mathematical game theory demonstrates that reasoning within classical first-order logic (CFOL) leads inevitably to the Nash equilibrium—a stable strategy profile in which no agent has an incentive to unilaterally deviate. While some Nash equilibria may coincide with Pareto-efficient outcomes, this is not the case under conditions of strategic uncertainty, as exemplified by the Prisoner’s Dilemma.

In that scenario, the incentive to betray—though individually rational—yields a Pareto-inefficient equilibrium: each co-conspirator defects, resulting in a collectively suboptimal outcome. The inefficiency does not arise from irrationality, but from the logical implications of strategic uncertainty under bounded communication and mutual distrust.

The key point is this: rational choice is not subjective. It follows deterministically from the logical system employed. Within CFOL, if agents operate with shared axioms and identical inference rules, they will inevitably arrive at the same conclusions. Disagreement can arise only if agents adopt different axiom sets—excluding trivial or mechanically detectable errors in algebra or arithmetic, which are operationally negligible in this context.

It is important to note that alternative logical systems—such as paraconsistent logic, modal logic, or intuitionistic frameworks—are formally excluded from the domain of classical game theory as currently defined. These systems may offer alternative conceptual models of reasoning, but they depart from the core assumptions of CFOL-based rationality, and are thus beyond the scope of this analysis.

Utility, Rationality, and the Bounds of Human Behavior

Rationality, though essential, is not a sufficient description of human action. At the foundation of both game theory and mathematical economics lies a deeper, more fundamental principle: a praxeological axiom, which holds that representative agents—or game-theoretic players—are not only rational, but also subjective utility maximizers. That is, agents act so as to maximize their payoffs, or more generally, their subjective welfare, referred to in economics as utility.

In mathematical economics, utility is commonly formalized through consumer and producer surplus, measured in monetary units as a standardized unit of account. A similar monetary interpretation often applies in mathematical game theory, particularly in models where payoffs are represented as quantifiable outcomes or incentives. While game theory may generalize utility to include non-monetary preferences, its core assumptions still presuppose agents seeking to optimize subjective valuation under constraints.

It is an empirically observed fact—without known counterexamples—that human agents consistently, systematically, and predictably pursue actions that increase their subjective utility, with rare exceptions such as ascetics or altruists. However, the stronger assumption of perfect rationality is clearly empirically untenable. A wide range of cognitive biases, including theory-induced blindness, loss aversion, and availability heuristics, systematically deviate from the idealized model of the rational agent.

Indeed, as early as the 1950s, Herbert A. Simon introduced the concept of bounded rationality, recognizing that agents operate under cognitive, informational, and computational constraints that limit their capacity for perfectly rational behavior. This framework concedes that while utility maximization remains a stable behavioral goal, the means of pursuing that goal are constrained by the architecture of human cognition and the limits of environmental information.

For this reason, any axiom of human behavior that aims to be empirically valid must account not only for the structural role of incentives and the cognitive limitations of agents, but also for the moral heterogeneity and ethical dispositions that differentiate individuals in both intention and outcome.

Axiom Zero: Formalizing Human Behavior Beyond Ideal Rationality

The proper formulation—the inviolate, logically consistent, and universally applicable praxeological axiom of human action—may be stated as follows:

Axiom Zero
Each representative agent is a boundedly rational, subjective utility maximizer who is susceptible to incentives. However, not all agents exhibit the same degree of ethical restraint or altruism—saints are rare outside of heaven.

This axiom reflects an empirical and formal truth: while all agents act in accordance with what they perceive to be in their own best interest, their belief systems, ethical constraints, and behavioral norms may differ dramatically. These differences are not matters of value judgment but of observable fact.

For example, the kamikaze pilot, willingly sacrificing his life for duty, and the agrarian Quaker, grounded in pacifism and simplicity, operate under distinct moral axioms—each coherent within its own normative system. Similarly, the behavioral incentives of a transnational gang member differ starkly from those of a devout monk or a public servant, yet all are agents in pursuit of utility as they define it.

The divergence in ethical priors and behavioral strategies among agents is not a defect in theory, but a foundational feature to be formally acknowledged. It reminds us that rationality is not a monolith, and that models of human action must account not only for instrumental optimization but for the heterogeneity of constraints, both internal and external.

This is not the domain of moral philosophy or rhetorical generalization. This is the domain of formal behavioral modeling.

Our task is not to moralize. It is to understand. Specifically, to define how bounded rationality manifests in practice—
Through formal structure, behavioral pattern, and model-theoretic representation.

We are not speaking of angels, but of agents.
Not of utopias, but of equilibria.

This is axiom zero, and every theory that hopes to touch reality must build from here.

Beyond the Obvious: Diagnosing the Roots of Bounded Rationality

Under the standard definition of perfect rationality in game theory, an agent is assumed to possess complete deductive power under classical first-order logic (CFOL): no ambiguity, no uncertainty, and no resource constraints. The agent applies inference rules flawlessly, deriving valid conclusions from a known and correct set of axioms.

Let us now set aside the obvious. We are not concerned with tasks like defeating a chess engine or factoring 2,000-digit primes—those are rightly outsourced to machines. Instead, we focus on decisions so simple and structurally transparent that a normal human—under no cognitive stress and operating from sound axioms—should reach the correct conclusion without difficulty. Think: the Pythagorean theorem, modus ponens, a short chain of inference. Middle-school math, at most.

For example:

• Understanding that switching doors in the Monty Hall problem is always the optimal strategy.

• Recognizing that active trading systematically underperforms the market, as demonstrated by Bill Sharpe in The Arithmetic of Active Management, and reiterated by John Bogle, Warren Buffett, and others.

And yet—many people fail to switch doors. Retail investors continue to engage in speculative trading.

Why?

If the logic is straightforward, the axioms are sound, and the conclusion is within reach, then what, exactly, does bounded rationality mean?

It means one of two things—and only two:

1. The agent misapplied the inference rules, or

2. The agent started from the wrong axioms.

That’s it.

• No appeal to complexity.

• No invocation of deep search trees, limited memory, or computational cost.

• Just a basic failure—either in deductive application or in foundational assumptions.

Bounded rationality, in nearly all real-world contexts, is not about being unable to beat a supercomputer at chess. It is, almost invariably, a matter of elementary error—disguised beneath layers of cognitive or institutional noise.

As we will show, most instances of bounded rationality do not arise from faulty reasoning. They arise from faulty premises—that is, from bad axioms. And here is the good news: by systematically auditing your axioms, you can eliminate nearly all bounded rationality in day-to-day decision-making.

What follows is not merely theoretical. It is a practical guide—a method for identifying and correcting the faulty assumptions that lead to theory-induced blindness.

Because that, ultimately, is what theory-induced blindness is:

A failure of axioms, not logic.

The most natural starting point for such an audit is to examine what the axioms actually claim. We begin, therefore, with the standard definition of utility in mathematical economics, as formalized by Arrow and Debreu.

Consumer and Producer Surplus: Integer-Valued Measures

Introduction
In mathematical economics—specifically in the Arrow–Debreu framework—welfare is formalized as a real-valued function:

  W(x):X→R
  where x∈X represents feasible allocations.

Within this structure, consumer and producer surplus are implicitly represented as real-valued differentials derived from utility functions—enabling proofs like the First and Second Welfare Theorems. This formulation holds in abstract equilibrium theory, where continuity, convexity, and differentiability are used to prove existence and optimality theorems.

However, this model introduces a representational gap when applied to real-world systems where surplus is not just an abstract preference delta, but a numerical value denominated in money—a fundamentally discrete quantity.

In all real-world economic systems, surplus exists in Z, not R, as money has a minimum unit (e.g., the cent). Representing it as a continuous variable assumes a level of precision that does not exist operationally. In formal logic—particularly in CFOL, where semantic precision is critical—this is a type mismatch. Transactions cannot occur at infinitesimal deltas of surplus. One cannot receive π cents.

This is not a critique of the Arrow–Debreu theorem itself, but of its naive application to systems requiring discrete precision—such as contracts, payments, or formal economic software. The BROSUMSI framework resolves this by redefining surplus as an integer-valued delta—ensuring logical alignment with both discrete economic actions and CFOL-based verification.

The BROSUMSI Framework

The BROSUMSI framework corrects this inconsistency by grounding all economic reasoning in three core principles:

• Bounded rationality
  Agents are not omniscient. They operate under cognitive constraints, using heuristics, local approximations, and selective attention.

• Subjective utility
  Every agent evaluates transactions based on its own internal utility gradient, ΔU—a private valuation shift per event.

• Integer accounting
  All surplus is calculated strictly over Z. No curves. No floating-point approximations. No mathematical fantasies.

In BROSUMSI, surplus is not the result of continuous optimization. It is a discrete, polarity-resolved transaction.
A pair of shoes has a subjective use-value and an objective exchange-value.
The surplus is the difference—in cents.

What Follows

What follows is a redefinition of consumer and producer surplus as integer-valued deltas under bounded cognition—formally encoded in Coq for verification.

No hand-waving.
No hidden assumptions.
Just logic. Structure. Proof.

Utility Under BROSUMSI

The BROSUMSI framework begins with a foundational praxeological axiom:

A representative agent opportunistically seeks to maximize their own subjective welfare,
operating within the constraints of bounded rationality, and while remaining susceptible to incentives.

Because BROSUMSI agents are, by definition, subjective utility maximizers, it is essential to define utility with precision and formal rigor.

Requirements for a Coherent Utility Definition

To ensure theoretical coherence and model interoperability, the definition of utility must satisfy the following conditions:

• ✅ Formally expressible within first-order logic (FOL)

• ✅ Consistent with standard formulations in mathematical economics and game theory

• ✅ Adapted to reflect the bounded, opportunistic, and incentive-sensitive behavior of BROSUMSI agents

Resulting Criteria

Within the BROSUMSI framework, a valid utility definition must be:

• Clear and unambiguous

• Logically consistent with FOL-based reasoning

• Interoperable with existing economic and game-theoretic models

—while also extending those models to incorporate the cognitive and epistemic limitations emphasized by the theory of bounded rationality.

Historical Foundations and Conceptual Background

The concept of utility traces its intellectual origins to the use–exchange value duality introduced by Aristotle nearly 2,500 years ago. According to Aristotle, every good or service—such as a pair of shoes—embodies two distinct aspects:

• Use Value: The good’s capacity to satisfy human needs
  (e.g., protecting one’s feet from injury)

• Exchange Value: The good’s monetary worth or market price

This dual structure was later adopted—though without explicit attribution—by Karl Marx in Das Kapital. Marx emphasized that a good (e.g., a winter coat) possesses:

• A use value (its ability to provide warmth)

• An exchange value (determined by prevailing market conditions)

Utility in BROSUMSI

Unlike traditional economic theories, which often treat utility as a cardinal, real-valued function, the BROSUMSI framework adopts a minimalist, ordinal foundation:

Utility is modeled as an integer-valued, ordinal function for decision-making purposes—nothing more.

This epistemically modest axiom bypasses philosophical debates over whether utility is metaphysically “real” or purely conceptual. Instead, it affirms a simple, pragmatic principle:

• If an agent is hungry and eats, their well-being improves.

• That improvement—whether in comfort, satisfaction, or happiness—is what BROSUMSI formally defines as utility.

Illustrative Example

The utility derived from wearing a warm coat may be both physical and social:

• Physical: It protects the body from cold and injury

• Social: A sable coat, for instance, may increase utility through:

  • Status signaling

  • Symbolic prestige

  • Cultural association

Compatibility with Existing Economic Models

This ordinal utility function remains fully compatible with standard economic frameworks, including:

• Consumer and producer surplus in microeconomics

• Welfare economics, particularly those focused on maximizing aggregate or average well-being

• Constitutional economics, especially normative models rooted in institutional or legal frameworks

Example: Pareto Efficiency and Welfare

Under standard interpretations of the U.S. Constitution’s directive to “promote the general welfare,” policy outcomes are often evaluated using Pareto efficiency:

An outcome is Pareto-efficient if no BROSUMSI agent can be made better off without making another worse off.

But worse off in what dimension?

In terms of subjective utility—as defined by each agent’s own ordinal welfare function.

🧠 Epistemic Framing

• Utility is not truth.

• It is a preference-ranking, expressed numerically to enable formal reasoning and decision logic.

• In BROSUMSI, utility is agent-relative, logic-compatible, and updateable over time.

Consumer and Producer Surplus: Integer-Valued Measures

Within the BROSUMSI framework, both buyers and sellers make decisions based on subjective cost–benefit reasoning, constrained by bounded rationality and cognitive limitations. As such, consumer and producer surplus are modeled not as cardinal utilities, but as integer-valued, ordinal utility differences, fully consistent with BROSUMSI’s foundational axioms.

Consumer Surplus (CS)

Let:

• P_max​(x) be the maximum price (in integer units) a consumer is willing to pay for good x, reflecting its subjective use value to the buyer.

• P(x) be the actual market price (in integer units), reflecting the objective exchange value agreed upon by buyer and seller.

Then:

CS(x)=P_max​(x)−P(x),where CS(x)∈Z

Interpretation:

• If CS(x)>0, the good is worth more to the buyer than its cost, and the transaction occurs.

• If CS(x)<0, the good costs more than it is subjectively worth, and no rational purchase is made.

Here, the subjective benefit is the agent’s internal valuation, the objective cost is the market price, and CS(x) represents the ex-ante expected utility of acquiring the good.

Producer Surplus (PS)

Let:

• P_min​(x) be the minimum price (in integer units) the seller is willing to accept, representing their subjective opportunity cost (e.g., disutility from effort, foregone leisure, or perceived risk).

• P(x) be the actual price received from the buyer (e.g., a wage or sale price).

Then:

PS(x)=P(x)−P_min​(x),where PS(x)∈Z

Interpretation:

• If PS(x)>0, the seller receives more than their subjective cost, and the transaction proceeds.

• If PS(x)<0, the seller incurs a net subjective loss, and the transaction is irrational.

A typical BROSUMSI agent trades labor or time for goods or compensation, making P_min​(x) analogous to a reservation wage, adjusted for disutility due to boredom, fatigue, hazard, or stress.

Surplus as Subjective Utility

Both CS(x) and PS(x) represent marginal improvements in subjective welfare. All values are:

• Agent-relative (evaluated from the agent's own point of view)

• Integer-valued (consistent with real-world accounting systems)

• Cognitively bounded (reflecting heuristic or non-optimal inference)

A voluntary exchange occurs if and only if both agents independently perceive that:

CS(x)>0 and PS(x)>0

This preserves the classical structure of mutual benefit, now reinterpreted through subjective utility gaps rather than idealized cardinal utility.

BROSUMSI Compatibility

Under the BROSUMSI axiom, each agent opportunistically seeks to maximize their own subjective welfare, within cognitive bounds and under the influence of incentives.

This welfare is operationalized as follows:

• For consumers, the utility metric is CS(x).

• For producers, it is PS(x).

These surplus measures are:

• Operational — requiring no metaphysical assumptions about utility.

• Bounded-rational — computed under constraints of limited inference, attention, and memory.

• Subjective — grounded in individual belief systems rather than objective optimization.

🧠 Key Insight

In BROSUMSI, surplus is not a measure of economic truth.
It is a function of agent-relative value, filtered through the lens of bounded cognition.

BROSUMSI Utility: An Ordinal Comparator-Based Approach

Formalizing Rational Choice Without Cardinal Illusions

1. Ordinal Decision-Making: The Comparator Function

Traditional utility theory relies on real-valued (cardinal) functions, which introduce inconsistencies under bounded rationality. BROSUMSI resolves this with a minimalist ordinal structure.

Formal Definition
Let X be a finite set of alternatives. For each agent i, define:

C_i : X × X → {+1, 0, -1}

Semantics
For any alternatives x, y ∈ X:

C_i(x, y) =

+1 if x ≻_i y (x is strictly preferred by i over y)

0 if x ~_i y (x is equivalent to y for agent i)

-1 if y ≻_i x (y is strictly preferred by i over x)

Example Scenario
An agent evaluating “10 cash” vs. “10 in Bitcoin” under a gain frame might have:

C_i(10 Bitcoin, 10 cash) = +1

meaning that the agent strictly prefers 10 in Bitcoin over 10 cash.

Key Advantages

• Context Sensitivity
  C_i(x, y) may reverse under different frames (e.g., gains vs. losses).

• Reference Dependence
  Captures endowment effects; for example: C_i(keep $10, lose $10) = +1
  reflecting the tendency to prefer keeping what one already has.

• Non-Additivity
  No assumption is made that: C_i(x, z) = C_i(x, y) + C_i(y, z)
  thereby avoiding contradictions in state-dependent preferences.

2. Eliminating Additive Inconsistencies

Cardinal linearity, as expressed by: U(a + b) = U(a) + U(b)

fails under bounded rationality.

Examples

• Satiety

C_i(eat 1st cookie, eat 2nd cookie) = +1

C_i(eat 10th cookie, stop) = -1

• Regret

Losing $100 followed by another $100 does not feel like twice the loss of $100. In other words: U(-100) and U(-200)
do not relate linearly—a violation of additivity that is naturally captured by BROSUMSI’s ordinal structure.

3. BROSUMSI’s Resolution

BROSUMSI uses pairwise comparisons between available strategies or consumption options. Let ΔU be an ex-ante utility difference between two alternatives x and y. We formalize the link between cardinal and ordinal logic by requiring:

∀ x, y ∈ X, sign(u_ex_ante_i(x) - u_ex_ante_i(y)) = C_i(x, y)

That is, the sign of the subjective utility difference determines the comparator output, and the magnitude |ΔU| is discarded—eliminating noise from arbitrary scaling.

4. Formalizing Nash Equilibrium in BROSUMSI

(a) Best Response Definition

Let S_i be the set of strategies available to agent i, and let:

C_i : S_i × S_i → {+1, 0, -1}

be agent i’s context-sensitive comparator over their own strategies, defined conditional on the current strategies of all other agents, denoted by s_{-i}. We write this as:

C_i(s_i′, s_i″) | s_{-i}

— meaning that agent i compares s_i′ versus s_i″ in the context where the other agents are playing s_{-i}. The best response set for agent i is defined as:

BR_i(s_{-i}) = { s_i′ ∈ S_i | ∀ s_i″ ∈ S_i, C_i(s_i′, s_i″) | s_{-i} ≠ -1 }

That is, s_i′ is not strictly dispreferred to any other available strategy under the current conditions—a maximal strategy under agent i’s ordinal preference.

(b) Nash Equilibrium Condition

A strategy profile s* = (s_1*, ..., s_n*) is a BROSUMSI Nash equilibrium if and only if:

∀ i ∈ N, s_i* ∈ BR_i(s_{-i}*)

This means that each agent is playing a strategy that is not strictly dominated by any alternative (given the strategies of others)—no agent prefers to deviate.

(c) Termination Condition for Equilibrium Computation

To compute equilibrium via best-response iteration, update the strategy profile as:

s(t+1) = (BR_1(s_{-1}(t)), ..., BR_n(s_{-n}(t)))

The process terminates when:

s(t+1) = s(t)

At that point, all agents are playing maximal strategies with respect to their current environment, and no further improvement is possible.

(d) Example: Prisoner’s Dilemma (Ordinal Form)

Let:

S_i = {cooperate, defect}

Define the comparator for each agent i (given the current strategy profile s_{-i}) as:

C_i(defect, cooperate) | s_{-i} = +1

C_i(defect, defect) | s_{-i} = 0

C_i(cooperate, defect) | s_{-i} = -1

Now suppose the current strategy profile is:

s* = (defect, defect)

Then, for each agent i, the best response is:

BR_i(defect) = { defect }

Thus, mutual defection is a BROSUMSI Nash equilibrium because:

C_i(cooperate, defect) = -1

implies that every deviation from defect results in a loss. No agent prefers to change strategies—even without the use of cardinal utilities.

5. Practical and Computational Advantages

Decision Rule

D(X) = argmax_{x ∈ X} Σ_{y ∈ X} C_i(x, y)

This selects the alternative x with the most pairwise “wins.”

Features

• Time Complexity: O(n log n) (e.g., achievable via merge sort)

• Space Complexity: O(n) — only comparisons are stored

• Noise Immunity: Robust to small shifts in ΔU (e.g., 1.0 → 1.1)

6. Behavioral Alignment and Empirical Consistency

BROSUMSI naturally encodes:

Loss Aversion

C_i(lose $100, gain $100) = -1

Framing Effects

For instance, under a loss-emphasizing frame:

C_i(save 200, let 400 die) = -1

Status Quo Bias

C_i(current job, new job) = 0

These outcomes match those of Prospect Theory—without the need for additional λ-parameters.

7. Formal Equivalence to Cardinal Utility

Step 1: Cardinal → Ordinal
Given a utility function u_i(x), define:

C_i(x, y) =

+1 if u_i(x) > u_i(y)

0 if u_i(x) = u_i(y)

-1 otherwise

Step 2: Ordinal → Cardinal
Given C_i, define:

u_i(x) = count of y ∈ X where C_i(x, y) = +1

Note: This mapping assumes that C_i is transitive. If not, tournament-style scoring may be employed.

Step 3: Why BROSUMSI Wins

Feature

Cardinal Utility

BROSUMSI Comparator

Requires cardinal magnitude

✅

❌

Supports framing effects

❌

✅

Time Complexity

O(n²)

O(n log n)

8. Computational Implementation (Scheme R7RS)

Pseudocode:

(define (comparator x y context)

(let ((delta (- (u-ex-ante x) (u-ex-ante y))))

(cond ((> delta 0) +1) ; x ≻_i y

((< delta 0) -1) ; y ≻_i x

(else 0)))) ; x ~_i y

(define (best-response strategies others context)

(car (sort strategies

(lambda (a b)

(> (comparator a others context)

(comparator b others context))))))

This implementation enables:

• Dynamic agent modeling

• Context tracking via continuations

• State-aware strategy selection

Optionally, the context can evolve over time as agents learn.

9. Formal Verification (Coq)

Lemma (comparator_antisymmetry):

Lemma comparator_antisymmetry :

forall x y, comparator x y = - comparator y x.

Proof.

(* Case analysis on x and y *)

intros x y; destruct (comparator x y) eqn:H;

destruct (comparator y x) eqn:H'; try reflexivity.

Qed.

Coq can be used to verify:

• Comparator antisymmetry (required)

• Comparator transitivity (optional)

• Strategy convergence

• Belief-preference alignment

• Nash equilibrium stability

10. Comparator Transitivity and Convergence Behavior

The BROSUMSI comparator C is antisymmetric:

• If C(x, y) = -1, then C(y, x) = +1

• If C(x, y) = 0, then C(y, x) = 0

However, transitivity is not required:

• C(x, y) = -1 and C(y, z) = -1 does not imply C(x, z) = -1

Thus, preference cycles are allowed:

x ≻_i y ≻_i z ≻_i x

Transitivity in BROSUMSI: Empirical, Not Axiomatic

• Transitive agents tend to converge to Nash equilibrium

• Non-transitive agents may exhibit cycling, instability, or bounded irrationality

• This is not a bug—it is a feature

Simulation Insight:

• Transitive agents lead to predictable dynamics

• Non-transitive agents introduce bounded rationality and noise

• Convergence depends on network-wide transitivity, not solely on individual logic

Coq Specification:

(* Required *)

Axiom comparator_antisymmetry :

forall x y : Alt, compare x y = - compare y x.

(* Optional (for rational agents) *)

Definition comparator_transitive (C : Alt -> Alt -> Z) :=

forall x y z : Alt,

C x y = -1 -> C y z = -1 -> C x z = -1.

Definition rational_agent (C : Alt -> Alt -> Z) :=

comparator_transitive C.

You can also define a NonTransitiveAgent type to explicitly model bounded irrationality.

11. Conclusion

Antisymmetry prevents contradictions.
Transitivity fosters convergence—but is not strictly required.

In BROSUMSI, non-transitivity is not a flaw—it is a feature that allows for the simulation of bounded rationality.
Agents who prefer A over B, B over C, but C over A are bounded, not broken.

Summary and Implications

BROSUMSI ordinal utility achieves:

1. Logical Coherence

2. Cognitive Realism

3. Formal Flexibility

4. Computational Tractability

In a world of limits, ordinal comparators don't simplify—they reveal.
And in BROSUMSI, they rule.

Surplus as Subjective Utility (BROSUMSI)

I. Core Premise

In the BROSUMSI framework, consumer and producer surplus are not approximated over real-valued utility curves. They are modeled as discrete, integer-valued deltas (ΔU) in subjective welfare, measured and resolved on a per-transaction basis. Every surplus is:

• Agent-relative: Defined by internal belief, not market consensus.

• Integer-valued: Formally measured over ℤ, not ℝ.

• Cognitively bounded: Based on local heuristics, not global optimization.

A transaction occurs if and only if both agents subjectively believe it will improve their welfare: CS(x) > 0 and PS(x) > 0

This reflects a bounded, epistemic interpretation of mutual benefit.

II. Transaction Anatomy

Each transaction is modeled as a pair of utility deltas — one per agent — with two time-indexed stages:

• du_ex_ante — expected utility at decision time

• du_ex_post — realized utility after resolution

Formal Structure

A transaction includes:

• buyer : agent purchasing the good

• seller : agent providing the good

• good : the object (good or service) being exchanged

• du_buyer : agent’s utility state (ex-ante and ex-post)

• du_seller : same, for the seller

In Coq:

Record du_per_t := {

du_ex_ante : Z;

du_ex_post : WorldState -> Z;

t_decision : Time;

t_resolve : Time

}.

Record Transaction := {

buyer : BROSUMSI_Agent;

seller : BROSUMSI_Agent;

good : Good;

du_buyer : du_per_t;

du_seller : du_per_t

}.

Before resolution, ex-post utility is epistemically gated:

Definition ex_post_du (du : du_per_t) (w : WorldState) : du_output :=

if time_le du.(t_resolve) (get_time w) then

Realized (du.(du_ex_post) w)

else

BraKet.

Matrix View

Each transaction yields a 4-value audit matrix:

(CS_ex_ante, PS_ex_ante, CS_ex_post, PS_ex_post)

Uses:

• Audit: Track mismatches between expected and realized outcomes

• Regret Analysis: Quantify decision error

• Systemic Diagnostics: Detect bias, asymmetry, or equilibrium failure

III. Failure Modes of Surplus

A. Epistemic Error

Agents act on belief, not knowledge. A voluntary trade can still yield loss:

du_ex_ante > 0, but du_ex_post < 0

Regret becomes formal:

Regret = du_ex_post – du_ex_ante

It’s not noise—it’s a structural mismatch between belief and outcome.

B. DIBIL — Dogma-Induced Blindness Impeding Literacy

Definition: An agent treats a belief as immutable, resisting update even in the face of contradiction.

Structure:

• du_ex_post < 0

• But du_ex_ante remains unchanged

Example:
A cult member donates everything, logs du_ex_ante = +100000, and after the outcome, still refuses to update despite du_ex_post = -100000.

This isn’t error. It’s bounded irrationality hardcoded into belief structure.

C. Strategic Exploitation

One agent knows the truth; the other holds a false belief.

Structure:

• du_ex_ante > 0 for the deceived

• du_ex_post < 0 after the trade

Examples:

• Snake oil

• MLM scams

• Religious tithes from epistemically isolated agents

This is valid structure, but predatory design.

D. Coercion

The agent acts under threat, not free choice. Though the transaction appears voluntary, it violates BROSUMSI’s core axiom.

Indicators:

• Agent cannot exit (e.g., monopoly, blackmail)

• No valid du_ex_ante exists

• Action is forced, not chosen

Result:
The ledger shows movement, but it wasn’t the agent who moved.

E. Systemic Misalignment

Institutions induce non-beneficial trades structurally.

Causes:

• Price distortion (e.g., rent control, monopsony)

• Legal or ideological prohibitions

• Market failure by design

Effect:
Trades occur with CS(x) = 0 or PS(x) = 0
Surplus is not realized—even if the transaction is allowed.

IV. Classifying Transactions in DeuxOS

Axes

• Voluntariness
   Voluntary — agent acts by free choice
   Involuntary — agent acts under coercion, constraint, or structural pressure

• Information Symmetry
   Symmetric — both agents have comparable epistemic access (knowledge, understanding, context)
   Asymmetric — one agent has significant epistemic advantage; the other acts from an incomplete or misled frame (as is inevitable when non-saints trade with bounded agents)

The 2×2 Grid

Symmetric Info

Asymmetric Info

Voluntary

Legitimate Trade

Exploitative Trade

Involuntary

Robbery (or Mutual Duress)

Slavery / Predatory Extraction

Definitions and Interpretation

• Legitimate Trade
   Both agents choose freely, based on accurate beliefs.
   → CS > 0 and PS > 0 are epistemically valid and mutually acknowledged.
   → Structural polarity is aligned.
   Example: Buying shoes at a fair, known price.

• Exploitative Trade
   Both agents appear to choose freely, but one is systematically misinformed or misled.
   → CS > 0 and PS > 0 are declared, but only one is real.
   → Asymmetric ledger truth.
   Example: Miracle cures, pump-and-dumps, cult donations.

• Robbery / Mutual Duress
   Neither agent truly consents.
   → Trade occurs under shared external pressure (e.g., war economy, hyperinflation, famine).
   → Epistemic conditions invalid; CS and PS are non-agentic.
   Example: Trading a gold ring for bread in a siege city.

• Slavery / Predatory Extraction
   One agent enforces an exchange under threat or total power asymmetry.
   → Trade is neither voluntary nor informed for one side.
   → The system records value transfer, but not consent.
   Example: Debt bondage, slave labor, monopoly-driven coercion.

🧠 DeuxOS Ledger Implication

In this schema, only Voluntary + Symmetric trades are truth-valid.

All other quadrants produce structural entropy:

• Asymmetry introduces epistemic misalignment.

• Involuntariness introduces agency suppression.

• Both distort ledger polarity, and accumulate as systemic drift or extraction.

However, the key conclusion, as sad as it is, is that under the axom that each representative agent is a boundedly rational, subjective utility maximizer who is susceptible to incentives — but not all agents exhibit the same degree of sainthood, outside of heaven, that is, it becomes self-evidently clear that the rent-seeking lemma of this axiom will hold true universally, as explained here.