Reflection principles and second-order choice principles with urelements (Q2668001)

In the paper under review, the author shows that the First-Order Reflection Principle (i.e., \(\forall x_{1},\ldots,x_{n}\forall Y_{1},\ldots,Y_{m}(\varphi\rightarrow\exists t(t\) is a transitive set \(\wedge\varphi^{t}))\), where \(\varphi\) contains no second-order quantifiers -- without the latter requirement on \(\varphi\), the above schema is the Second-Order Reflection Principle) is not provable in Kelley-Morse set theory with urelements (it is not assumed in this theory that the urelements form a set), denoted by KMU, with Global Choice (i.e., ``There is a function \(F\) such that for every non-empty set \(x\), \(F(x)\in x\)''). This is accomplished by constructing a suitable model (satisfying Set Choice) with a \textit{proper class} of urelements. In particular, the author works in KMU + Limitation of Size (i.e., KMU + ``For any two proper classes \(X\) and \(Y\), there exists a bijection \(F:X\rightarrow Y\)'') with Ur being a countably infinite set of urelements in the universe \(U\). Let \(\vartriangleleft\) be a dense linear ordering of Ur without first or last elements, and let \(\mathcal{G}\) be the group of permutations on Ur which preserve \(\vartriangleleft\). A class \(X\) is symmetric if and only if there is a finite set \(s\subseteq\mathrm{Ur}\) such that \(\phi(X)=X\) for all \(\phi\in Fix(s)=\{\pi\in\mathcal{G}:\forall u\in s(\pi(u)=u)\}\). Let \(\mathcal{M}\) be the permutation model determined by the hereditarily symmetric classes, i.e., those classes \(X\) which are symmetric and all elements of the transitive closure \(TC(X)\) of \(X\) are symmetric also. Let \(\mathcal{N}\) be the model which is constructed as follows: \(x\) is a set in \(\mathcal{N}\) if and only if \(x\) is a set in \(\mathcal{M}\) and \(TC(x)\cap\mathrm{Ur}\) (the kernel of \(x\)) is finite. \(X\) is a class in \(\mathcal{N}\) if and only if \(X\) is a class in \(\mathcal{M}\) and every element of \(X\) is either a set in \(\mathcal{N}\) or a urelement. The author shows that \(\mathcal{N}\) satisfies KMU\(^{+}\) + Global Choice, where KMU\(^{+}\) denotes KMU with a proper class of urelements, but not First-Order Reflection. (The above model was originally used in [\textit{U. Felgner}, Stud. Logic Found. Math. 84, 217--255 (1976; Zbl 0341.02054)] and [\textit{P. E. Howard} et al., J. Symb. Log. 43, 673--684 (1978; Zbl 0442.03032)] to show that Global Choice does not imply Global Well-Ordering: There is a well-ordering of \(U\).) The following results are also established in the paper: \begin{itemize} \item[1.] KMU\(^{+}\) + Second-Order Reflection + Limitation of Size and KM + Second-Order Reflection are mutually interpretable. \item[2.] KMU\(^{+}\) + Limitation of Size and KM are bi-interpretable with parameters. \item[3.] Assuming the consistency of KM + \(\exists \kappa\)(\(\kappa\) is a \(\kappa^{+}\)-supercompact cardinal), KMU\(^{+}\) + Second-Order Reflection + \(\neg\)(Limitation of Size) is consistent. \end{itemize} The paper is concluded with the following open question: What is the optimal consistency strength of KMU\(^{+}\) + Second-Order Reflection + \(\neg\)(Limitation of Size)?