[Paper Review] On safe post-selection for Bell nonlocality: Causal diagram approach

Paweł Błasiak, Ewa Borsuk+1 more|arXiv (Cornell University)|Dec 14, 2020

Philosophy and History of Science4 citations

TL;DR

This paper introduces the 'all-but-one principle'—a causal diagram-based criterion—to determine when post-selected data in Bell nonlocality experiments can safely be used to infer genuine nonlocality. By modeling spatially distributed post-selection using causal inference tools, the method ensures that conclusions about nonlocality remain valid despite selection bias, providing a general framework for analyzing multipartite Bell inequalities across diverse entanglement generation schemes.

ABSTRACT

Reasoning about Bell nonlocality from the correlations observed in post-selected data is always a matter of concern. This is because conditioning on the outcomes is a source of non-causal correlations, known as a selection bias, rising doubts whether the conclusion concerns the actual causal process or maybe it is just an effect of processing the data. Yet, even in the idealised case without detection inefficiencies, post-selection is an integral part of every experimental design, not least because it is a part of the entanglement generation process itself. In this paper we discuss a broad class of scenarios with post-selection on multiple spatially distributed outcomes. A simple criterion is worked out, called the all-but-one principle, showing when the conclusions about nonlocality from breaking Bell inequalities with post-selected data remain in force. Generality of this result, attained by adopting the high-level diagrammatic tools of causal inference, provides safe grounds for systematic reasoning based on the standard form of multipartite Bell inequalities in a wide array of entanglement generation schemes without worrying about the dangers of selection bias.

Motivation & Objective

• To address concerns about selection bias in post-selected Bell nonlocality experiments, where conditioning on outcomes may distort causal inferences.
• To develop a general criterion that identifies when conclusions about nonlocality remain valid despite post-selection on spatially distributed outcomes.
• To provide a systematic, causally sound method for analyzing multipartite Bell inequalities in real-world entanglement generation schemes involving post-selection.
• To formalize the conditions under which post-selection does not invalidate nonlocality conclusions, even in idealized scenarios without detection inefficiencies.

Proposed method

• Employing high-level causal diagram tools from causal inference to model post-selection as a conditioning process on multiple spatially distributed outcomes.
• Defining the 'all-but-one principle' as a sufficient condition under which post-selection does not introduce spurious nonlocal correlations.
• Analyzing the structure of causal graphs to identify when the observed correlations in post-selected data reflect genuine nonlocality rather than selection bias.
• Applying the framework to a broad class of multipartite Bell inequality scenarios, including those arising in common entanglement generation protocols.
• Using structural causal modeling to distinguish between genuine causal influences and spurious correlations induced by post-selection.
• Deriving conditions under which the violation of Bell inequalities in post-selected data can be trusted as evidence of nonlocality.

Experimental results

Research questions

• RQ1 Under what conditions can post-selected data in Bell nonlocality experiments still provide valid evidence of nonlocality?
• RQ2 How can selection bias introduced by post-selection on multiple spatially distributed outcomes be formally ruled out as a confounding factor?
• RQ3 What general criterion ensures that a violation of a multipartite Bell inequality in post-selected data reflects genuine nonlocality rather than spurious correlations?
• RQ4 In what types of entanglement generation schemes can post-selection be safely used without invalidating nonlocality conclusions?
• RQ5 How can causal inference tools be systematically applied to validate nonlocality claims in post-selected experimental data?

Key findings

• The all-but-one principle provides a sufficient condition under which post-selection on multiple spatially distributed outcomes does not invalidate conclusions about nonlocality.
• The criterion ensures that observed violations of Bell inequalities in post-selected data can be trusted as evidence of genuine nonlocality, even in the absence of detection inefficiencies.
• The method enables safe analysis of nonlocality in a wide range of multipartite entanglement generation schemes where post-selection is inherent to the protocol.
• Causal diagram modeling successfully isolates the effects of post-selection, distinguishing them from genuine nonlocal correlations.
• The framework applies generally to standard forms of multipartite Bell inequalities, offering a reusable and systematic approach for experimental validation.
• The approach removes the need to worry about selection bias in post-selection when the all-but-one condition is satisfied, enhancing reliability in experimental design and interpretation.