- If an event E has high probability, accept regularity as an explanation; otherwise move to the next step.
- If the chance hypothesis assigns E a high probability or E is not specified, then accept chance; otherwise move down the list.
- Having eliminated regularity and chance, accept design.

This filter is equivalent to detecting complex specified information.

Dembski, William A., 1998. *The Design Inference: Eliminating chance
through small probabilities*. Cambridge University Press.

- The filter is useless in practice because the probabilities it asks for
can never be known. Step 1, in particular, does not ask us to accept
or reject just one regularity hypothesis, it asks us about all
regulatory hypotheses, even ones that nobody has thought of before.
Similarly, rejecting chance requires a complete list of all chance
processes that might apply to the event.
- The filter is based on the premise that the categories of regularity,
chance, and design are "mutually exclusive and exhaustive" (Dembski
1998, 36). But they are not mutually exclusive. R. A. Fisher, for
example, included mutations in all three categories. Individually,
they were due to chance, but collectively they were governed by laws,
and all of this was planned by God (Ruse 2001, 121).
- Although the filter claims to detect design, it really says nothing
about design. The filter defines design as the elimination of
regularity and chance, not, as most people would define design, as
purposeful, intelligent arrangement. The two definitions are not
equivalent. Dembski himself noted that some intelligent design will be
eliminated in the first two steps. And what the filter actually
detects is copying, not intelligent
agency.
- Since the filter does not say anything about design, there is no
intelligent design hypothesis that can be used scientifically or for
any practical purposes.
- Key terms in the filter, especially "chance hypothesis" and
"specified," are poorly defined.
- Dembski does not consider that design is a process. The process that produces design is itself not regularity (or the resulting design would have high probability) or chance (or the design would likely not result), so the filter says the process must itself be design. Thus, the design process must have another design process to produce it, which needs a design process of its own, ad infinitum, or somewhere along the way there must be no process at all and design must come out of nowhere. In actuality, design is typically done as an iterative process involving lots of trial and error. Regularity and chance are both parts of the process, as is selection. Evolution uses the same processes.

Wilkins, John S. and Wesley R. Elsberry, 2001. The advantages of theft over toil: the design inference and arguing from ignorance.

- Ruse, Michael, 2001.
*Can a Darwinian Be a Christian?*, Cambridge University Press.

Pennock, Robert T., 1999.

created 2003-2-18, modified 2003-11-8