A situation where $n$ voters choose between two candidates can be modelized by a $n$-uple $(a_1,\ldots,a_n)\in\lbrace 0,1 \rbrace^n$, where $a_i$ denotes the wish of the $i$-th voter. So the voting system (the rule which decides which candidate will be elected from the uple) can be described as a boolean map $f:\lbrace 0,1 \rbrace^n \to \lbrace 0,1 \rbrace$. The following assumptions are natural on $f$ :
(1) Symmetry : $f(1-a_1,1-a_2,\ldots,1-a_n)=1-f(a_1,\ldots,a_n)$ for any uple $a_1,\ldots,a_n$.
(2) Monotonicity : if $a_k\leq b_k$ for every $k$, then $f(a_1,\ldots,a_n) \leq f(b_1,\ldots,b_n)$.
If $I\subsetneq \lbrace 1,2,\ldots, n \rbrace$ is such that $f$ does not depend on the coordinates $a_j$ for $j\not\in I$, we say that $I$ is a (nontrivial) coalition ; the trivial coalition is $I=\lbrace 1,2,\ldots, n \rbrace$. When there is no nontrivial coalition, we say that $f$ is irreducible.
The question is, what are the irreducible maps $f:\lbrace 0,1 \rbrace^n \to \lbrace 0,1 \rbrace$ satisfying (1) and (2) ? When $n$ is odd, there is an obvious solution of "choosing the candidate with the most votes". I couldn't find any other examples.
My thoughts :
This is very closely related to Arrow's theorem and similar results, of course. But everything I've found so far only dealt with voting systems with at least three outcomes.
Update: The question can also be expressed in "measure-theoretic" terms. Indeed, if for $A\subsetneq \lbrace 1,2,\ldots, n \rbrace$ we denote by $i_A$ the indicator uple of $A$ (i.e. the uple $u=(u_1,\ldots,u_n)$ with $u_a=1$ when $a\in A$ and $0$ otherwise) and put $\mu(A)=f(i_A)$, then for the "measure" $\mu$, (1) means that $\mu(A)\leq\mu(B)$ when $A\subseteq B$, and (2) means that $\mu(A^{c})=1-\mu(A)$ where $A^{c}$ denotes the complement of $A$.