Experimental test of measurement dependent local Bell inequality with human free will

TL;DR

This paper reports an experimental test of a measurement dependent Bell inequality using human-generated random numbers, demonstrating violation that constrains local hidden variable models with limited measurement independence.

Contribution

It provides the first detailed experimental validation of measurement dependent Bell inequalities with human choices, addressing the measurement independence assumption.

Findings

Bell inequality violation confirmed with human randomness

Measurement independence parameter constrained to l > 0.10 ± 0.05

Supports the nonlocal nature of quantum correlations with human-influenced settings

Abstract

A Bell test can rule out local realistic models, and has potential applications in communications and information tasks. For example, a Bell inequality violation can certify the presence of intrinsic randomness in measurement outcomes, which then can be used to generate unconditional randomness. A Bell test requires, however, measurements that are chosen independently of other physical variables in the test, as would be the case if the measurement settings were themselves unconditionally random. This situation seems to create a "bootstrapping problem" that was recently addressed in The BIG Bell Test, a collection of Bell tests and related tests using human setting choices. Here we report in detail our experimental methods and results within the BIG Bell Test. We perform a experimental test of a special type of Bell inequality - the measurement dependent local inequality. With this inequality, even a small amount of measurement independence makes it possible to disprove local realistic models. The experiment uses human-generated random numbers to select the measurement settings in real time, and implements the measurement setting with space-like separation from the distant measurement. The experimental result shows a Bell inequality violation that cannot be explained by local hidden variable models with independence parameter (as defined in [Putz et al. Phys. Rev. Lett. 113, 190402 (2014).] ) l > 0.10 +/- 0.05. This result quantifies the degree to which a hidden variable model would need to constrain human choices, if it is to reproduce the experimental results.