Retrieving the Baby: Reichenbach's Principle, Bell Locality, and Selection Bias

This paper argues that the failure of Bell's Factorizability condition arises from selection bias rather than a violation of intuitive locality, suggesting that EPR-Bell correlations are a selection artifact akin to collider bias that does not require abandoning Reichenbach's Principle of Common Cause.

The Gist

The Big Idea: "We Didn't Throw the Baby Out; We Just Missed It"

Imagine you are washing a baby in a bathtub. You are scrubbing so hard to get the dirt (the confusing math of quantum physics) out that you accidentally throw the baby (a crucial piece of logic) out with the dirty water.

This is what the famous physicist John Bell warned us about in the 1990s. He showed that quantum particles seem to be "spooky" and connected across vast distances, defying our normal understanding of how the world works. He called this Nonlocality.

But Bell also warned that in his attempt to turn his "intuitive" ideas into strict math, he might have thrown something important away.

Huw Price's paper argues that Bell was right to be suspicious. He didn't just throw away a baby; he threw away a specific kind of baby: Selection Bias.

Price argues that the "spooky" connection between quantum particles isn't because they are magically talking to each other faster than light. Instead, it's because we are looking at a biased sample of data. It's a statistical trick, not a magical one.

---

1. The Problem: The "Spooky" Connection

In the quantum world, if you take two particles that are "entangled" (linked together) and send them to opposite sides of the universe, they seem to coordinate their behavior instantly. If you measure one, the other instantly "knows" what to do.

Standard physics says this violates Locality (the idea that things can only be influenced by their immediate surroundings). To explain this, we usually have to accept one of two scary things:

1. Magic: The particles are communicating instantly across space (violating Einstein's speed limit).
2. Pre-determination: The universe was rigged from the beginning to make it look like they are communicating (Superdeterminism).

Bell's math (called Factorizability) says: "If you know the past causes, the two particles should act independently." But experiments show they don't act independently. So, Bell concluded, the universe must be non-local.

2. The Solution: The "Survivorship Bias" Analogy

Price says Bell missed a third option: Selection Bias.

To understand this, let's look at a famous story from World War II.

The Bomber Plane Story:
Imagine you are a statistician in WWII. You look at the planes that return from bombing missions. You see bullet holes all over the wings and the tail, but zero bullet holes in the engines.

• The Wrong Conclusion: "The engines are tough! We don't need to reinforce them. We should reinforce the wings."
• The Right Conclusion (Abraham Wald): "The planes that got hit in the engines didn't come back. We are only looking at the survivors. The data is biased because the dead planes aren't in our sample."

This is Survivorship Bias. The correlation (holes in wings vs. no holes in engines) isn't because the wings are special; it's because we filtered out the data that didn't survive.

Price's Argument:
Quantum experiments are doing the exact same thing.
When we measure entangled particles, we aren't seeing all the possible outcomes. We are only seeing the ones that "survived" a specific selection process. The "spooky" correlation appears only because we are looking at a filtered sub-group of data, not the whole picture.

3. The "W-Shape" Experiment: The Magic Filter

Price uses a specific quantum experiment (called Entanglement Swapping) to prove his point. Imagine a "W" shape in time:

• Two pairs of particles are created.
• One particle from each pair goes to a central station (the middle of the W).
• The other two go to Alice and Bob (the ends of the W).

Here is the magic trick:

1. Before the middle measurement: Alice and Bob's particles are not correlated. They are just random noise.
2. The Middle Measurement: A scientist at the center measures the two middle particles.
3. The Filter: Based on the result of that middle measurement, the scientist discards some data and keeps other data.
4. The Result: Suddenly, looking only at the kept data, Alice and Bob's particles look perfectly correlated!

The Analogy:
Imagine Alice and Bob are flipping coins in separate rooms. Usually, their results are random.

• A "Referee" in the middle watches the coins.
• The Referee says: "If I see a specific pattern in the middle, I will keep the results from Alice and Bob. If I see a different pattern, I will throw their results in the trash."
• When you look only at the results the Referee kept, Alice and Bob's coins seem to be magically linked.

But they aren't linked! The link is an illusion created by the Referee's filter. The "spooky" connection is just a Selection Artefact.

4. Why This Matters: Saving "Locality"

If Price is right, we don't need to believe in "spooky action at a distance."

• The "Baby" we saved: The idea that the universe is Local (things only affect their neighbors).
• The "Bathwater" we threw out: The idea that every correlation must be explained by a common cause in the past (Reichenbach's Principle).

Price argues that in the quantum world, correlations can arise simply because of how we select the data (like the Referee in the W-experiment). We don't need to assume the particles are talking to each other. We just need to admit that our experiment is filtering the data in a way that creates a fake pattern.

5. The "Time-Symmetry" Twist

There is one catch. In our daily life, we usually think of causes happening before effects. But Price suggests that in the quantum world, we might need to be more flexible with time.

He suggests that if we look at the universe without our usual bias toward "past causes future," the math works out perfectly. The "filter" (the selection) can happen in a way that looks like it's happening in the future, but it explains the data without breaking the speed of light.

Summary: The Takeaway

• The Myth: Quantum particles are magically connected across space.
• The Reality: They aren't connected. The "connection" is an optical illusion caused by Selection Bias.
• The Analogy: It's like looking at a photo of a crowd where only the people wearing red hats are visible. You might think "Everyone in this crowd loves red hats!" But really, you just took a photo that only showed people with red hats. The correlation is in the photo, not the crowd.
• The Result: We can keep our belief that the universe is local (no faster-than-light magic) if we accept that quantum experiments are essentially "filtering" the data in a way that creates these spooky patterns.

In short: Bell didn't throw the baby out; he just missed the baby hiding in the bathwater. The baby is Selection Bias, and once you find it, the "spooky" quantum world becomes much less spooky and much more logical.

Technical Summary

1. The Problem

The paper addresses the foundational tension in quantum mechanics (QM) between Bell's Theorem and the principle of Locality.

• The Conflict: Bell's Theorem demonstrates that the predictions of QM (confirmed by experiment) violate Factorizability (FACT). FACT is a mathematical rule derived from an intuitive notion of "local causality" (PLC-1), which states that events in spacelike separated regions should not influence each other directly.
• The Dilemma: Standard interpretations conclude that QM requires "nonlocality" (spacelike influence) or the abandonment of Reichenbach's Principle of Common Cause (PCC). PCC asserts that correlations must be explained by either direct causation or a common cause that "screens off" the correlation.
• Bell's Warning: John Bell himself warned that the mathematical derivation of FACT from intuitive locality might involve "throwing the baby out with the bathwater," suggesting that a crucial conceptual element might have been lost in the formalization.
• The Gap: While some have argued EPR-Bell correlations are exceptions to PCC, they have not been successfully categorized under existing, well-understood exceptions. The paper seeks to identify the "missing baby" Bell alluded to.

2. Methodology

Price employs a combination of causal modeling, statistical analysis, and conceptual reconstruction:

• Re-evaluating PCC: The author revisits Reichenbach's Principle, specifically its known exceptions, focusing on Selection Bias (including survivorship bias and collider bias).
• Minimal Selection Model (MSM): A framework is proposed where correlations differ between a "super-ensemble" (the total set of possible cases) and a "sub-ensemble" (the selected cases). If a selection process creates correlations in the sub-ensemble that do not exist in the super-ensemble, the correlation is a "selection artefact."
• Toy Models:
  • Classical Postselection: A model where a third party (Charlie) filters classical coin-toss outcomes based on quantum probabilities to reproduce Bell correlations.
  • Quantum Preselection: A model where the initial Bell state is chosen randomly, creating a super-ensemble with no Bell correlations, which are then revealed only in sub-ensembles defined by the specific state.
  • W-shaped Experiments: Analysis of Entanglement Swapping protocols (where two independent pairs are entangled via a Bell-state measurement). Here, Bell correlations between distant particles (A and B) emerge only after postselecting on the measurement outcome at a central node (M).
• Time-Symmetry Conjecture (TSC): The author invokes a time-symmetric regime (removing the "Past Hypothesis" of low entropy) to argue that the statistical structure of preselection in standard Bell tests is equivalent to the postselection in W-shaped experiments, removing the need for "gerrymandered" initial conditions.

3. Key Contributions

• Identification of the "Baby": The paper identifies Selection Bias as the "baby" Bell inadvertently discarded. It argues that EPR-Bell correlations are not a unique violation of PCC but fall under the well-known exclusion of Selection Artefacts.
• Reframing Bell Correlations: The author proposes that Bell correlations are selection artefacts (specifically akin to collider bias or range restriction). They arise because the experimental data represents a selected sub-ensemble of a larger, uncorrelated (or differently correlated) super-ensemble.
• Distinction between Ontology and Statistics: The paper clarifies that labeling correlations as "artefacts" does not deny the reality of entanglement. Rather, it diagnoses the origin of the statistical correlation as a selection effect, thereby relieving the pressure to explain them via direct spacelike causation or common causes.
• Time-Symmetric Causality: The paper suggests that in a time-symmetric regime (without the Past Hypothesis), the counterfactual structures required to define "colliders" exist in all Bell experiment geometries (V-shaped and W-shaped), not just in Delayed Choice Entanglement Swapping.

4. Results

• Failure of Factorizability (FACT) is Explained: The failure of FACT is shown to be a natural consequence of selection bias. When conditioning on a selection variable (the measurement outcome or the specific entangled state), correlations appear between variables that are independent in the broader ensemble.
• Resolution of the Locality Paradox:
  • PLC-1 (Intuitive Locality): Preserved. The proposal requires no violation of relativistic locality (no spacelike influence).
  • PLC-2 (Local Causality as defined by Bell): Failed. However, this failure is attributed to the inapplicability of PCC to selection artefacts, not to nonlocality.
  • FACT: Failed. But again, this is due to the selection process, not hidden nonlocal variables.
• No Need for New Exemptions: EPR-Bell correlations do not require a novel exception to PCC; they fit into the existing category of selection bias (collider bias).
• Independence from Measurement Independence (SI): The proposal does not require challenging the assumption of Statistical Independence (SI) via superdeterminism or retrocausality in the traditional sense. It simply posits that the observed correlations are selection effects, which is compatible with standard operational QM.

5. Significance

• For Quantum Foundations: The paper offers a "local" explanation for Bell correlations that avoids the need for spooky action-at-a-distance. It suggests that the "nonlocality" of QM is an artifact of how data is selected and conditioned, rather than a fundamental feature of spacetime causality.
• For Causal Discovery and Statistics: It highlights the ubiquity of selection bias in quantum phenomena. It warns causal modelers that applying standard PCC-based reasoning to quantum data without accounting for selection effects (collider bias) leads to false conclusions about nonlocality.
• Re-evaluating Bell's Legacy: It vindicates Bell's suspicion that the transition from intuitive locality to mathematical Factorizability was flawed. The "bathwater" was the intuitive concept of locality; the "baby" was the statistical reality of selection bias, which Bell missed.
• Philosophical Implications: It challenges the necessity of time-asymmetry in causal reasoning at the fundamental level. By invoking the Time-Symmetry Conjecture, it suggests that a time-symmetric ontology can naturally accommodate the operational features of QM without violating relativistic constraints.

In summary, Price argues that Bell nonlocality is a statistical illusion caused by selection bias. By recognizing Bell correlations as selection artefacts, one can maintain a locally causal universe (PLC-1) while acknowledging that the mathematical conditions for local causality (PLC-2 and FACT) fail because they ignore the role of selection in generating correlations.