MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Say that a preordering $\le$ on a set of subsets of some space preserves strict inclusion provided that $A\lt B$ whenever $A\subset B$ (where $A\lt B$ iff $A\le B$ and $B \not\le A$).

Let the space be the unit circle $T$. Then it's easy to see that there is no preorder on all countable subsets of $T$ that preserves strict inclusion and that satisfies the condition that $A \le B$ iff $r[A] \le B$ for every rotation $r$. (Just let $A = \{ e^{2\pi i n u} : n \in \mathbb N \}$ for an irrational number $u$, and let $r$ be a rotation by $-2\pi u$.)

Is there a preordering of all Borel-measurable subsets of $T$ that (a) preserves strict inclusion, (b) satisfies the condition that $A \le B$ iff $r[A] \le r[B]$ for every rotation $r$ and (c) is total, i.e., all Borel subsets are comparable? (Without (c), we could let $\le$ be $\subseteq$.)

(One can think of a total preorder as a very general generalization of measure or probability, so the question is whether there is any measure-like thing that preserves strict inclusion, i.e., is "regular" in Bayesian terminology.)

Update: The answer is negative if we require $r[A] \le r[B]$ for all isometries $r$ of the circle (i.e., we include reflections). That's because (b)-for-isometries and (c) in that case imply that $A \le B$ iff $r[A] \le B$ for every isometry $r$, and we already saw that in that case we have a problem.

To see that $A \le B$ iff $r[A] \le B$, note that all rotations are compositions of reflections, and so all we need to show is that if $A \le B$, then $\sigma[A] \le B$ for any reflection $\sigma$. If $\sigma$ is a reflection and $A \le B$, suppose for a contradiction that we don't have $\sigma[A] \le B$. Then by (c) we have $\sigma[A] > B$. By the analogue of (b), we have $A = \sigma^2[A] > \sigma[B]$. Since $B \ge A$, we have $B > \sigma[B]$. Thus, by (c) $\sigma[B] > \sigma^2[B]$ and so $\sigma[B] > B$, and we have a contradiction.

But I still don't know what happens if instead of all isometries we just have all rotations.

share|cite|improve this question
What if instead of T we ask the same question about Z? The integers are a subgroup of the circle, so it should be easier to give a preorder on them, if there is one. – domotorp Sep 8 '13 at 8:23
Indeed: a negative answer for Z will give a negative answer for T. – Alexander Pruss Sep 9 '13 at 2:21
But in fact we have a positive answer for T, hence also for Z. :-) – Alexander Pruss Sep 16 '13 at 4:39
up vote 1 down vote accepted

The answer is positive, given Choice. It turns out that whenever $G$ is an abelian group acting on a set $X$, then there is a $G$-invariant preorder $\le$ on $2^X$ such that if $A$ is a proper subset of $B$, then $A<B$.

The proof uses the fact that $G$-invariant partial orders on $\Omega$ extend to $G$-invariant linear orders, provided that $G$ is abelian and no element of $G$ has a finite orbit of length greater than one. While $2^X$ in general won't satisfy the orbit condition, one can instead take $\Omega$ to be a quotient of $2^X$ by the relation $A\sim B$ iff there is a $g\in G$ and $n\in\mathbb N$ such that $gB=A$ and $g^nB=B$.

A writeup is here.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.