A decision-theoretic approach to dealing with uncertainty in quantum mechanics

This paper proposes a decision-theoretic framework that derives Born's rule from utility functions associated with quantum measurements, thereby uncoupling precise probability theory from quantum mechanics to accommodate imprecise probabilities while providing a foundation for recent work by Benavoli, Facchini, and Zaffalon and offering a distinct alternative to the approaches of Deutsch and Wallace.

The Gist

The Big Picture: A New Way to Look at Quantum Guessing

Imagine you are playing a high-stakes game of poker against a mysterious opponent. In the quantum world, this opponent is a tiny particle (like an electron), and the cards it holds are its "quantum state."

Usually, when physicists try to predict what this particle will do, they use probability. They say, "There is a 50% chance the particle will spin up, and a 50% chance it will spin down." This is the famous Born's Rule.

However, this paper asks a fundamental question: Why do we have to use precise probabilities? Why can't we just say, "I'm pretty sure it's spin up, but I'm not 100% sure"?

The authors (Keano De Vos, Gert De Cooman, Alexander Erreygers, and Jasper De Bock) propose a new way to think about this. Instead of starting with math that forces us to have exact numbers, they start with decisions. They argue that we can understand quantum uncertainty by looking at what a rational person would choose to do, without assuming we know the exact odds beforehand.

The Core Idea: Betting on Measurements

To explain their theory, the authors use a simple setup:

1. You (The Player): You are uncertain about the state of a quantum system.
2. The Act: You can choose to perform a specific measurement (like checking if a spin is up or down).
3. The Reward: If you measure the system, you get a "reward" (like money or points) based on the outcome.

In standard quantum mechanics, the reward is calculated using a strict formula (the Born Rule). The authors ask: Can we derive this formula just by looking at how a rational person makes decisions?

They say yes, but with a twist. They don't assume you have to be able to rank every single possible outcome perfectly. You might be undecided between two options. This is where they introduce Imprecise Probabilities.

The Analogy: The "Fuzzy" Map vs. The "Perfect" Map

Think of your knowledge about the quantum system as a map.

• The Old Way (Standard Quantum Mechanics): The map is perfectly detailed. It tells you exactly where you are and exactly what will happen next. It leaves no room for doubt. If you have this map, you can always say, "I prefer Option A over Option B."
• The New Way (This Paper): The map is a bit fuzzy. You know you are in a certain region, but you aren't sure of the exact coordinates. Because of this fuzziness, you might look at two paths and say, "I can't decide which is better right now."

The authors show that it is perfectly rational to have this "fuzzy" map. You don't need to force a decision if you don't have enough information.

The Four Rules of the Game

To make their theory work, the authors set up four simple rules (postulates) that any rational player should follow. These rules are like the laws of physics for decision-making:

1. The Certainty Rule: If you know for a fact that a measurement will give a specific result (say, +1), then the value of that measurement is exactly +1. No guessing needed.
2. The "Same Game, Different Room" Rule: If you play a game in one room (Hilbert space) and an identical game in another room, the value of the game should be the same. The physical location doesn't change the math.
3. The Additivity Rule: If you combine two measurements, the total value is the sum of their individual values. (If Game A is worth 5 points and Game B is worth 3, doing both is worth 8).
4. The Smoothness Rule: If you make a tiny change to the system, the value of the measurement shouldn't jump wildly. It should change smoothly.

The Magic Result: Born's Rule Appears Naturally

Here is the "magic trick" of the paper.

The authors start with these four simple decision rules and the idea that you might be uncertain (fuzzy map). They don't start with the Born Rule. They don't even start with the idea of "probability."

They run the math, and poof! The Born Rule pops out as a special case.

• If you are totally uncertain: You end up with a "set" of possible probabilities (a range of possibilities). This is the Imprecise Probability approach. It's like saying, "The chance is somewhere between 40% and 60%."
• If you happen to know the exact state: The "fuzzy" range collapses into a single, precise number. Suddenly, you get the standard Born Rule (e.g., "The chance is exactly 50%").

The Analogy: Imagine you are trying to guess the temperature.

• Imprecise approach: You look out the window and say, "It's probably between 60 and 70 degrees."
• Precise approach: You walk outside with a thermometer and say, "It is exactly 65 degrees."
• The Paper's Point: The "thermometer reading" (precise probability) is just a special, very specific case of the "looking out the window" (imprecise probability) approach. You don't need to assume the thermometer exists from the start; it emerges naturally when you have perfect information.

Why This Matters

The authors compare their work to two famous scientists, Deutsch and Wallace, who tried to prove the Born Rule using decision theory.

• Deutsch and Wallace assumed that you must be able to rank every single option perfectly (a "total ordering"). They assumed you always know exactly what you prefer.
• The Authors say: "No, that's too strong." In real life, we often can't decide between two things if we don't have enough info. By allowing for indecision (partial ordering), their theory is more flexible and realistic.

They show that you can still get the standard quantum rules (Born's Rule) even if you allow for indecision. In fact, allowing for indecision gives you a more powerful toolbox to handle situations where we simply don't know enough.

The "Heisenberg vs. Schrödinger" Connection

The paper also mentions a cool mathematical symmetry. In quantum mechanics, there are two ways to describe how a system changes:

1. Heisenberg Picture: You focus on the measurements (the tools you use).
2. Schrödinger Picture: You focus on the state (the object you are measuring).

The authors show that their "decision theory" approach naturally connects these two pictures.

• Thinking about "Desirable Measurements" (what you want to do) is like the Heisenberg picture.
• Thinking about "Sets of Density Operators" (the mathematical representation of your uncertainty) is like the Schrödinger picture.
• Their math proves these two ways of thinking are actually two sides of the same coin.

Summary

This paper argues that quantum mechanics doesn't force us to use precise probabilities.

Instead, it suggests that:

1. We should start with decisions (what we prefer to do).
2. We should allow for uncertainty and indecision (we don't always have to pick a winner).
3. If we do this, the famous Born's Rule (the standard quantum probability formula) appears naturally as a special case when we happen to have perfect information.

It's a way of saying that the "weirdness" of quantum mechanics isn't about magic probabilities, but about the logical structure of how we make choices when we don't know the full story.

Technical Summary

1. Problem Statement

The paper addresses the foundational issue of how to model uncertainty in quantum mechanics (QM). Traditionally, QM relies on precise probability theory (specifically, density operators and Born's rule) to describe both:

1. Epistemic uncertainty: Uncertainty about which quantum state a system is in (mixed states).
2. Intrinsic quantum uncertainty: Uncertainty about measurement outcomes even when the state is known.

The authors argue that the standard approach imposes a "total ordering" on preferences, forcing agents to have precise probabilities for all outcomes. This leaves no room for indecision, imprecision, or partial information. The paper seeks to:

• Decouple the mathematical structure of QM from the assumption of precise probabilities.
• Provide a decision-theoretic foundation that allows for imprecise probabilities (sets of probabilities) in quantum contexts.
• Derive Born's rule not as a fundamental postulate, but as a special case arising from more general decision-theoretic principles.

2. Methodology

The authors adopt a decision-theoretic framework inspired by de Finetti and the theory of imprecise probabilities (specifically coherent sets of desirable gambles), rather than the Savage framework used by Deutsch and Wallace.

Core Setup

• Acts: Measurements ($\hat{A}$) on a quantum system are treated as "acts" or "gambles."
• States: The quantum state $|\Psi\rangle$ is the unknown variable (the "state of the world").
• Rewards: The outcome of a measurement is mapped to a real-valued reward (utility) expressed in "utiles."
• Preferences: The agent (You) expresses a strict partial preference ordering ($\triangleright$) over these uncertain rewards. Crucially, this ordering is not required to be total; the agent may be unable to compare two acts (incomparability) or may simply lack a preference (indecision).

Key Postulates

The authors introduce four postulates regarding the reward function $w_{\hat{A}}(|\psi\rangle)$, which maps a measurement $\hat{A}$ and a state $|\psi\rangle$ to a real reward:

1. RF1 (Eigenstate Certainty): If the system is in an eigenstate of $\hat{A}$ with eigenvalue $\lambda$, the reward is exactly $\lambda$.
2. RF2 (Invariance): The reward depends only on the eigenvalues and the superposition weights, not on the specific Hilbert space embedding or basis labeling. This ensures invariance under unitary transformations and relabeling.
3. RF3 (Additivity): For commuting operators with the same eigenspaces, the reward function is additive ($w_{\hat{A}+\hat{B}} = w_{\hat{A}} + w_{\hat{B}}$).
4. RF4 (Continuity): The reward function is continuous with respect to the state.

3. Key Contributions and Results

A. Derivation of the Reward Function (The Main Theorem)

The central mathematical result (Theorem 13) proves that the postulates RF1–RF4 uniquely determine the reward function.

• Result: The reward for performing measurement $\hat{A}$ on state $|\psi\rangle$ is necessarily:
  $$w_{\hat{A}}(|\psi\rangle) = \langle \psi | \hat{A} | \psi \rangle$$
• Significance: This recovers the standard quantum mechanical expectation value formula. Crucially, this is derived without assuming Born's rule or the existence of probabilities a priori. It emerges as the unique "fair price" (coherent prevision) consistent with the decision-theoretic axioms.

B. Generalization to Imprecise Probabilities

The paper extends this framework to cases where the state $|\Psi\rangle$ is unknown (epistemic uncertainty).

• Coherent Sets of Desirable Measurements: Instead of a single probability distribution, the agent's beliefs are represented by a coherent set of desirable measurements ($D$). A measurement $\hat{A}$ is in $D$ if the agent strictly prefers it to the status quo (zero reward).
• Coherent Lower Previsions: These sets are mathematically equivalent to coherent lower previsions ($\underline{P}$), which assign a lower bound to the expected reward of any measurement.
• Credal Sets of Density Operators: The authors prove that any coherent lower prevision corresponds to a convex closed set of density operators ($\mathcal{R}$).
  • The lower prevision is the minimum expectation over this set: $\underline{P}(\hat{A}) = \min_{\rho \in \mathcal{R}} \text{Tr}(\rho \hat{A})$.
  • Recovery of Standard QM: If the set $\mathcal{R}$ collapses to a singleton (a single density operator), the model reduces to standard quantum mechanics with precise probabilities.

C. Comparison with Deutsch and Wallace

The paper contrasts its approach with the decision-theoretic derivations of Born's rule by Deutsch and Wallace:

• Deutsch/Wallace: Assume the system is in a known state and require a total ordering of preferences. They aim to justify probabilities in a Many-Worlds (Everettian) context.
• De Vos et al.: Allow for unknown states and partial orderings (incomparability).
  • They argue that the "totality" requirement is an unnecessary constraint that hides the conservative nature of inference.
  • Their approach is more general: it encompasses the Deutsch/Wallace results as a special case (when the state is known and preferences are total) but allows for imprecise models when information is incomplete.

D. Heisenberg vs. Schrödinger Duality

The framework naturally recovers the duality between the Heisenberg and Schrödinger pictures:

• Heisenberg Picture: Representing uncertainty via sets of measurements (desirable gambles).
• Schrödinger Picture: Representing uncertainty via sets of density operators (states).
  The paper shows these are mathematically equivalent representations within their decision-theoretic framework.

4. Significance and Implications

1. Probabilities are Derivative: The paper argues that probabilities (and density operators) are not fundamental to quantum mechanics but are derivative tools that emerge from a more fundamental decision-theoretic structure involving preferences and rewards.
2. Handling Partial Information: By allowing for imprecise probabilities (sets of density operators), the framework provides a rigorous way to handle situations where an agent has limited information about a quantum system, without forcing arbitrary precise probabilities.
3. Conservative Inference: The approach utilizes conservative inference (deductive closure). If an agent only knows that a measurement is "desirable," the framework allows them to infer the minimal set of other measurements that must also be desirable, without over-committing to specific probabilities.
4. Computational Utility: The authors suggest that imprecise models can be computationally advantageous. In complex systems, calculating exact expectations might be intractable; calculating conservative bounds (lower/upper previsions) via sets of density operators can make inference feasible (e.g., via "lumping" techniques).
5. Foundational Shift: It offers a path to justify quantum mechanics without relying on the "measurement problem" or specific interpretations (like Many-Worlds) to derive probabilities. Instead, it grounds the Born rule in the rationality of an agent dealing with uncertain rewards.

Conclusion

The paper successfully constructs a decision-theoretic foundation for quantum mechanics that generalizes the standard formalism. It demonstrates that Born's rule is a special case of a broader theory of imprecise probabilities applied to quantum measurements. By treating measurements as acts with uncertain rewards and allowing for partial preference orderings, the authors provide a robust framework for reasoning about quantum uncertainty that is both mathematically rigorous and philosophically distinct from previous decision-theoretic derivations.