Contextuality from Single-State Ontological Models: An Information-Theoretic No-Go Theorem

This paper establishes an information-theoretic no-go theorem proving that classical ontological models constrained to reuse a single ontic state space across multiple interventions inevitably incur an irreducible contextual information cost, thereby identifying contextuality as a fundamental limitation of such classical representations that quantum theory circumvents by relaxing the single-variable assumption.

The Gist

Here is an explanation of the paper using simple language, everyday analogies, and metaphors.

The Big Idea: The "One Backpack" Problem

Imagine you are a traveler with a single, fixed backpack (this is the "ontic state"). You are going on a trip where you have to visit three different cities: City A, City B, and City C.

In each city, you have to follow a specific set of rules to get a "score" (this is the "measurement outcome").

• In City A, the rules say: "If you have a red hat, you get 10 points."
• In City B, the rules say: "If you have a red hat, you get 0 points."
• In City C, the rules say: "If you have a red hat, you get 5 points."

The Catch: You cannot change your backpack. You cannot swap it for a "City A Backpack" or a "City B Backpack." You must carry the same backpack into every city, and the contents of that backpack (the "ontic state") must stay exactly the same.

The Problem: The "Context" Gap

The paper argues that if you try to explain your scores using only what is inside your single backpack, you will hit a wall.

In a normal, classical world, if you know exactly what's in your backpack, you should be able to predict your score no matter which city you visit. But in this specific scenario (which mimics how quantum particles behave), the rules of the game are tricky:

• If you look at City A and City B together, the rules make sense.
• If you look at City B and City C together, the rules make sense.
• But if you try to write down a single "master plan" for your backpack that works for all three cities at once, it becomes mathematically impossible.

The paper calls this Contextuality. It means your score depends not just on what you have (the backpack), but on where you are (the context/city).

The New Discovery: The "Information Tax"

The author, Song-Ju Kim, proves a new "No-Go Theorem." Here is the translation:

The Old Way of Thinking:
Scientists used to think, "Maybe the backpack is just too small! If we make the backpack huge and fill it with more details, we can explain everything."

The New Finding:
The author says: No, size doesn't matter. Even if your backpack is the size of a warehouse and contains infinite information, you still cannot explain the scores using only the backpack.

Why?
Because the "rules" of the cities (the interventions) interact with the backpack in a way that requires extra information that isn't inside the backpack.

• To explain why you got 10 points in City A but 0 points in City B, you need to know which city you are in.
• But the backpack doesn't know which city you are in (because the backpack is the same in all cities).
• Therefore, you have to carry a second piece of information (a "Context Ticket") just to tell the backpack which rules to apply.

The paper calls this an "Irreducible Information Cost." It's a tax you have to pay. You cannot simulate this behavior classically without adding this extra "Context Ticket."

The Quantum Twist: How Nature Cheats

So, how does the real world (Quantum Mechanics) do it without paying this tax?

The paper explains that nature doesn't use a single backpack.

In Quantum Theory, when you go to City A, the universe doesn't just check your backpack. It essentially creates a new version of reality specifically for City A. When you go to City B, it creates a new version for City B.

• It doesn't try to force one single "backpack" to explain all three cities simultaneously.
• It relaxes the rule that "everything must come from one single underlying source."

Because quantum mechanics allows for this flexibility (it doesn't force a single global probability space), it avoids the "Information Tax" entirely. It doesn't need the extra "Context Ticket" because it doesn't try to cram incompatible rules into one fixed box.

A Simple Analogy: The Chameleon

Imagine a chameleon (the system) sitting on a leaf.

• Classical View (The Paper's Constraint): You insist the chameleon has a fixed skin pattern (the ontic state) that never changes. You ask, "What color will the chameleon be?"
  • If the chameleon is on a green leaf, it turns green.
  • If the chameleon is on a red leaf, it turns red.
  • The Problem: If you insist the chameleon has one fixed pattern that determines its color, you can't explain why it changes color based on the leaf. You have to admit that the "leaf" (the context) is providing extra information that the "skin" (the state) doesn't have.
• Quantum View: The chameleon doesn't have a fixed pattern. Its "skin" is a superposition of possibilities. The act of looking at it on a specific leaf creates the color. It doesn't need to carry extra info because the color is generated in the moment by the interaction, not pre-written in the skin.

Why Should You Care? (Implications for AI and Minds)

The paper suggests this isn't just about physics; it applies to Artificial Intelligence and human thinking.

Imagine a robot (or a human brain) with limited memory. It has to make decisions in many different situations (contexts) using the same internal memory (the "single backpack").

• If the robot tries to act "rationally" using only its fixed memory, it will fail in complex, changing environments.
• It will need to constantly carry "extra baggage" (contextual info) to make sense of the world.
• The paper suggests that non-classical ways of thinking (like quantum-inspired AI) might be better for these systems because they don't force everything into one rigid, single-state box. They allow for a more fluid way of handling context.

Summary in One Sentence

You cannot explain how a system reacts differently to different situations using only a single, fixed internal state; you are forced to pay an "information tax" by adding extra context, unless you abandon the idea of a single fixed state entirely (which is what Quantum Mechanics does).

Technical Summary

Here is a detailed technical summary of the paper "Contextuality from Single-State Ontological Models: An Information-Theoretic No-Go Theorem" by Song-Ju Kim.

1. Problem Statement

The paper addresses the fundamental nature of contextuality in quantum theory. Traditionally, contextuality is understood as the impossibility of reproducing quantum measurement statistics using noncontextual ontological models (models where measurement outcomes are determined solely by an underlying physical state, independent of the measurement context).

While existing no-go theorems (e.g., Kochen-Specker) establish this impossibility based on algebraic structures or specific quantum formalisms, the information-theoretic origin of this constraint remains less understood. Specifically, the paper investigates whether contextuality arises as a fundamental limitation on classical representations that are constrained to reuse a single ontic state space across multiple interventions (contexts) without refining or indexing that space based on the context.

Core Question: Can a classical probabilistic model, constrained to a fixed, single ontic state space $\Lambda$ reused across all interventions, account for contextual dependence without incurring additional information-theoretic costs?

2. Methodology and Framework

The author employs a representation-theoretic and information-theoretic approach, independent of specific physical dynamics or quantum formalism.

• Single-State Ontological Models: The paper defines a strict class of classical models where:
  1. There is a fixed ontic state space $\Lambda$.
  2. This space is reused across all interventions (contexts) $C$.
  3. The space is not indexed, duplicated, or refined based on the intervention $C$.
  4. All statistics arise from a single underlying classical probability space over $\Lambda$.
• Interventions: Contexts are modeled as interventions acting on $\Lambda$ or the response functions, but the identity of the intervention is not explicitly encoded within $\Lambda$ itself.
• Information-Theoretic Tools: The analysis utilizes Shannon entropy and conditional mutual information to quantify the "cost" of representing contextuality.
  • $C$: Intervention (Context)
  • $\lambda$: Ontic state
  • $O$: Observable outcome
  • $M$: Auxiliary contextual variable (if required)

3. Key Contributions

The paper makes three primary theoretical contributions:

1. Definition of Single-State Constraints: It formalizes the concept of "single-state reuse," distinguishing it from standard hidden variable models that might allow context-dependent refinements. It isolates the constraint of information reuse within a fixed probability space.
2. Information-Theoretic No-Go Theorem: It proves that any classical ontological model satisfying the single-state constraint must incur an irreducible contextual information cost.
3. Reframing Contextuality: It shifts the explanation of contextuality from a structural property of quantum observables to a fundamental representational limitation of classical probability theory when forced to reuse a single state space.

4. Main Results

Theorem 1: The Information-Theoretic No-Go Theorem

The central result states that for any classical ontological model reusing a single ontic state space $\Lambda$ across interventions $C$, if the observable statistics exhibit contextual dependence, then:
$$H(M) \ge I(C; O | \lambda) > 0$$
Where:

• $I(C; O | \lambda)$ is the conditional mutual information between the intervention $C$ and the outcome $O$, given the ontic state $\lambda$.
• $H(M)$ is the entropy of an auxiliary contextual variable $M$ required to mediate the dependence.

Implication: Contextual dependence cannot be fully mediated through the ontic state $\lambda$ alone. The inequality $I(C; O | \lambda) > 0$ proves that knowing the ontic state is insufficient to predict the outcome; the specific context $C$ provides additional information not contained in $\lambda$. Therefore, a classical model must introduce an external variable $M$ (contextual information) to resolve the statistics, violating the "single-state" purity.

Constructive Illustration

The paper provides a constructive example (Section IV) demonstrating this obstruction:

• Consider three interventions with binary outcomes where pairwise marginals are consistent, but no global joint distribution exists (a standard signature of contextuality).
• Under the single-state constraint, reproducing these statistics requires encoding which marginal structure is active.
• A deterministic example shows that if $O = \lambda \oplus f(C)$, then $I(\lambda; O) = 0$ but $I(C; O | \lambda) = H(f(C)) > 0$. This explicitly quantifies the information cost required to bridge the gap between the state and the context-dependent outcome.

Why Quantum Theory Avoids the Obstruction

The paper explains that quantum theory avoids this no-go theorem because it relaxes the assumption that all measurement statistics arise from a single underlying classical ontic variable.

• In quantum mechanics, outcomes are generated via the Born rule ($p(o|C) = \text{Tr}(\rho E_o(C))$).
• While a single quantum state $\rho$ is reused, the outcomes are not assumed to arise from a single global classical probability space over ontic states that simultaneously reproduces all contexts.
• Thus, quantum theory does not require an auxiliary variable $M$ to mediate contextuality because it does not force the embedding of all contexts into a single classical probability space.

5. Significance and Implications

• Foundational Physics: The result provides a new perspective on the Kochen-Specker theorem. It suggests that contextuality is not merely a quirk of quantum algebra but a fundamental limitation on classical representation arising from the inability to reuse a single information space for multiple, incompatible contexts without cost.
• Adaptive Intelligence and AI: The paper draws parallels to bounded rationality and resource-limited systems (e.g., cognitive science, working memory).
  • Systems with fixed internal representations (single-state reuse) interacting with multiple contexts will inevitably exhibit "non-classical" statistical behaviors (order effects, context sensitivity).
  • These behaviors are not anomalies but necessary consequences of the information-theoretic cost of representing contextuality within a fixed memory capacity.
• Device Independence: The theorem is device-independent; it relies solely on the structural requirement of single-state reuse, not on specific physical implementations or dynamics.

Conclusion

Song-Ju Kim establishes that contextuality is an information-theoretic obstruction inherent to classical models constrained to reuse a single ontic state space. To reproduce contextual statistics, such models must pay an irreducible cost in the form of additional contextual information ($I(C; O | \lambda) > 0$). Quantum theory circumvents this by abandoning the requirement of a single global classical probability space, highlighting contextuality as a fundamental trade-off between adaptability and representational economy.