An Algorithmic Information Theory Critique of Statistical Arguments for Intelligent Design

TL;DR

This paper critiques W. Dembski's statistical and decision-theoretic arguments for intelligent design, showing they are flawed and do not provide scientific evidence against natural processes or evolution.

Contribution

It offers an algorithmic information theory critique of Dembski's decision process and clarifies misconceptions about randomness, natural laws, and information in the context of intelligent design.

Findings

Dembski's decision process is unworkable and flawed.

Universal randomness tests show natural world outcomes are highly ordered.

The so-called law of conservation of information is equivalent to the second law of thermodynamics.

Abstract

In a number of books and articles including "The Design Inference" and "No Free Lunch", W. Dembski claims to have established a robust decision process that can determine when observed structures in the natural world can be attributed to design. Dembski's decision process first asks whether a structure as an outcome can be explained by the regularity of natural laws. If not, and the outcome can be "specified", a randomness test is devised to determine whether an observed low probability outcome indicates design. It is argued in this paper that the Dembski test is unworkable and is better formulated in terms of a Martin Loef universal randomness test. A universal randomness test will show that most observed outcomes in the natural world are non random; they are highly ordered. However this does not necessarily demonstrate design, as the decision is not between chance and design, but between natural laws and design. The Dembski decision template, which eliminates natural processes in the first decision step, is flawed, forcing a design outcome when none is warranted. Dembski also introduces a 4th law of thermodynamics, the law of conservation of information to argue information cannot emerge from random processes. However if a more robust measure of information based on algorithmic entropy is used, the so called 4th law is seen to contain no more than the second law of thermodynamics. In conclusion, despite the good intentions, the approach fails to offer any new insights into the adequacy of evolutionary theory and should not be regarded as demonstrable science.