Relational hidden variables and non-locality (Q1956376)

This paper develops a relational framework, rather than a probabilistic model, to discuss the problem of hidden variables and non-locality. The essential formalism is the form \[ e: M\times O \rightarrow \{ 0,1\} \] where \(M\) is the set of measurements on the system and \(O\) the set of outcomes. If the result is 1 for a combination then this is a possible event in the system. The notation e denotes empirical since the elements of the relation can be observed. To include hidden variables an object \[ h \subseteq M \times O \times \Lambda \] is considered where \( \Lambda \) is the set of hidden variables. The definition of e realising a model is that for any \( (m,o) \in e \) we can find \( \lambda \in \Lambda \) where \[ h(m,o,\lambda) =1 \] This is a brief summary of the formalism which may oversimplify it slightly. It is very elegant but perhaps Prof. \textit{J. S. Bell} may not have liked the term ``measurement'' to describe the elements of M. (see [``Against `measurement'''. NATO Sci. Ser. B, Adv. Sci. Inst. Ser. 226, 17--31 (1990; \url{doi:10.1007/978-1-4684-8771-8_3})]). The paper takes the reader through a range of results where probability models can be described by relational models. One interesting one is the Hardy paradox where it is shown that in a particular system non-locality must be present and the results cannot be explained by hidden variables [\textit{L. Hardy}, Phys. Rev. Lett. 71, No. 11, 1665--1668 (1993; Zbl 0972.81522)]. In this paper, this is demonstrated using a relational model which provides a good example of the formalism: \[ M= M_{1} \times M_{2} = \{ (X_{1},Y_{1}),(X_{1},Y_{2}),(X_{2},Y_{1}),(X_{2},Y_{2}) \} \] \( X_{i}, Y_{i} \) being `measurements' on two quantities and the outcomes are described by \[ O = \{ (R,R),(R,G),(G,R),(G,G) \} . \] In the final discussion in the paper, the computational aspects are considered and it is shown that membership of the a class of hidden variables models is a NP problem.