Consider the example of a scantron multiple choice test. Jack totally forgot to study for the test, and so his only recourse was to fill in the blanks at random. Jill, however, studied hard and was able to select the correct answers. We can say that Jill is more intelligent than Jack. While Jack relied totally on chance, Jill was able to curb chance and select the right answers, in order to achieve the goal of getting an A on the test.
So in order to detect artifacts of intelligence, we have to find places in nature where chance was purposefully constrained. This is perhaps where Dembski's explanatory filter comes to play (i.e. if not regularity, then chance; if not chance, then design). We might have to modify the filter a little: if not regularity, then unconstrained chance; if not unconstrained chance, then constrained chance (= design).
Note that intelligence does not need to be the antithesis of chance; rather, intelligence can be the constraining of chance. An intelligent agent can sometimes use constrained chance as a useful tool (e.g. the sorting algorithm known as Quicksort).