When one is first learning about topologies or $\sigma$-algebras, a common (trivial) exercise is showing that they are closed under intersections, but not in general under unions, i.e., any intersection of topologies is a topology, but the union of two topologies may not be a topology. When I was first learning topology, I found this to be rather disappointing, for otherwise one could form the "topology of topologies" on a given set $X$.
Suppose we have a property $P$, such as "$\forall y,z\in x \:(y\cap z\in x)$." Now for any set $x$, we may form the set $K_P (x)$ of all $t\subset x$ such that $P(t)$ holds. Can we say anything about $P$ if, for all $x$, $P(K_P (x))$ holds? Of course, a trivial example of such a $P$ is the property of being closed under intersections. Is that the only such $P$ ?