What actually appears in TDI is a deductive argument for
finding the attribute of "complexity-specification" or
"complex specified information" in an event (ch. 2,3,4,5,6),
and an inductive argument that finding such an attribute
usually means that an intelligent agent has acted (section
2.4), thus the attribute always means that an intelligent
agent has acted. But the A-number-one objection that is sure
to be raised is not addressed by Dembski in TDI, which is that
natural selection is perfectly capable of producing events
with the "complexity-specification" attribute, which would
make the inductive argument for going from "design" (in the
sense Dembski uses it) to "agency" far less than a sure thing.
It will not be enough to show that the operation of natural
selection does not always produce CSI; it is going to require
that it be shown that NS can *never* produce CSI to move the
inductive argument forward.
Dembski has some useful insights on the elimination of chance
explanations, but I find a big mismatch between the claims of
the "First Things" article and what TDI actually delivers.
Maybe the solution is to ignore the "First Things" article if
one wants to appreciate TDI.
Wesley