Sure wins, separating probabilities and the representation of linear functionals (Q1018312)

For some class \({\mathcal A}\) of sets let \(P:{\mathcal A}\to R\) be a set function such that, for \(E\in{\mathcal A}\), \(P(E)\) is the price for betting 1 dollar on the occurrence of \(E\). In a paper of \textit{B. de Finetti} [Ann. Inst. Henri Poincaré 7, 1--68 (1937; Zbl 0017.07602)] \(P\) is called coherent if it admits no sure wins. De Finetti showed that if \({\mathcal A}\) is a finite algebra, then \(P\): \({\mathcal A}\to R\) is coherent iff \(P\) is a probability measure. This was extended and generalized e.g. by \textit{V. S. Borkar, V. R. Konda} and \textit{S. K. Mitter} [Stat. Probab. Lett. 66, No. 4, 417--421 (2004; Zbl 1081.28004)]. In the present paper, the author examines the absence of sure wins for a convex cone \({\mathcal K}\) of real-valued functions on some arbitrary set \(\Omega\), obtaining conditions for the existence of a finitely additive probability measure \(m\) such that \(\sup_{k\in{\mathcal K}}m(k)\leq 0\). If \({\mathcal K}\) is the kernel of some linear functional, this leads to the characterization of those functionals that admit the representation as finitely additive expectations. (This topic is also addressed in \textit{P. Berti} and \textit{P. Rigo} [Proc. Amer. Math. Soc. 128, No. 11, 3251--3258 (2000; Zbl 0952.28009)].)