Elephants Sold Separately: Understanding the Design Inference
My previous article here at ENV, "Pink EleP(T|H)ants on Parade," was written in response to blogger Elizabeth Liddle's post, "The eleP(T|H)ant in the room." Now, she has written another reply, "Trojan EleP(T|H)ant?" The exchange with Dr. Liddle is part of an ongoing discussion about the design inference, as developed by William Dembski. I return to the subject of Liddle because I think her misunderstandings of Dembski's work are shared by others. Some clarification is thus in order.
Well, it seems that I mistook the point that Liddle was trying to make. The design inference essentially seeks to prove a conditional: that if the development of life is highly improbable under any Darwinian scenario, then we are justified in inferring design. Initially, I understood Liddle to be objecting to the conditional. However, her objection is not to the logic of the conditional, but to the antecedent: namely that life is highly improbable given the Darwinist scenario.
She objects that the design inference does not show that Darwinism is improbable. Rather, she believes, it merely assumes that Darwinism is improbable. She points out that the probability of evolution is much more difficult to determine than the probability involved in rolling dice or tossing a coin. That is why she thinks there is an elephant in the room that we are ignoring. It is also why she has accused me of trying to sneak in a Trojan elephant.
("EleP(T|H)ant," incidentally, refers to her objection that we cannot calculate the P(T|H), the "Probability that we would observe the Target (i.e. a member of the specified subset of patterns) given the null Hypothesis." See my earlier articles for more information.)
But as I mentioned, the design inference is a conditional. It argues that we can infer design from the improbability of Darwinian mechanisms. If offers no argument that Darwinian mechanisms are in fact improbable. When proving a conditional, we are not concerned with whether or not the antecedent is true. We are interested in whether the consequent follows from the antecedent.
For that reason, in The Design Inference, Dembski offers no argument that biology is improbable by Darwinian mechanisms. He mentions the controversy over evolution, but the book is agnostic as to whether or not Darwinian evolution's probability of being true is high or low. In No Free Lunch, Dembski does argue that the development of biological complexity is improbable under Darwinism, but not by appealing to the design inference or specified complexity. He never merely assumes that Darwinism is an improbable account. Instead he appeals to the displacement problem and irreducible complexity.
Liddle and others who think that Darwinism provides a probable account of biological complexity are under no obligation to accept design. As long as they hold that the antecedent of the conditional is false, they are free to reject the consequent, the inference to design. If we want to determine whether or not Darwinism provides a probable account of origins, we need to look at other arguments and evidence, such as irreducible complexity, protein folds, or the conservation of information. Only in conjunction with these other arguments does the design inference allow us to draw a conclusion.
Liddle rejects the arguments put forward by intelligent design proponents to establish the low probability of Darwinian mechanisms. I do not think that her counterarguments work, but to show why would require a separate discussion.
In both of my last two articles, I emphasized that the design inference makes use of multiple hypotheses. It is not enough to reject a single possible explanation, based on chance and necessity, for a given object. Other possible explanations besides design could exist. It is necessary to be able to reject multiple hypotheses. In fact, the design inference depends on our being confident that the rejected hypotheses cover any and all chance- and necessity-based explanations that might be true. Only then can we conclude that the object was designed.
Liddle objects that I'm making this up. She asserts that Dembski has only ever worked with a single hypothesis. But that is simply not true. The formulation of the design inference has involved multiple hypotheses since Dembski first developed it.
On p. 222 of The Design Inference, Dembski presents the argument of the design inference, which begins: "Suppose a subject S has identified all the relevant chance hypotheses, H, that could be responsible for some event E."
On p. 72 of No Free Lunch, Dembski presents the Generic Chance Elimination Argument. Step 2 states: "S determines that only those chance hypotheses in the set [of relevant chance hypotheses] could have been operating to produce E." Step 8 states: "S is warranted in inferring that E did not occur according to any of the chance hypotheses ... and therefore that E exhibits specified complexity."
On p. 26 of his 2005 paper, "Specification: the pattern that signifies intelligence," Dembski refers to "the relevant collection of chance hypotheses that we have good reason to think were operating if the event E happened by chance is some collection of chance hypotheses." He then discusses how to "eliminate all these chance hypotheses." Liddle herself even quotes Dembski discussing the elimination of multiple propositions (essentially hypotheses): "Provided that the proposition along with its competitors form a mutually exclusive and exhaustive class, eliminating all the competitors entails that the proposition is true."
It is thus abundantly clear that multiple hypotheses have always been the basis for the design inference.
In rejecting individual hypotheses, Dembski employs the criterion of specified complexity. "Specification" means that an object conforms to a pattern independent of the object itself. It is the difference between Mt. Rushmore and a random mountainside. "Complexity" refers to the probability of an object arising by chance, and must be calculated according to the hypothesis under consideration.
Liddle's account of these ideas seems inconsistent. In her post "Specification for Dummies," she computes the complexity of an image of a glacier being generated entirely at random and not according to any relevant hypothesis. She states: "So, by Dembski's definition, my pattern is Complex (has lots of Shannon Bits)." She goes on to say: "Dembski has a third hurdle that my pattern has to vault." And she explains the pattern must also be highly improbable under the chance hypothesis.
In "The eleP(T|H)ant in the room," she defines complexity as: "One of a very large number of patterns that could be made from the same elements (Shannon Complexity)."
In both of those posts, Liddle appears to be calculating complexity in term of simple randomness regardless of what the hypothesis under consideration is. She then goes on to consider the probability under the chance hypothesis as if it were some third criterion, in addition to complexity and specification. That is not quite accurate.
However, in her most recent post, Liddle does indicate that the chance hypothesis should be taken into account when computing complexity. This is correct. It just does not seem to be the same as she was saying before.
I hope that I have helped clarify matters, for Elizabeth Liddle and others. She objects that the method of considering multiple hypotheses is an innovation, but it's easy to show that Dembski has been writing about multiple hypotheses since he first presented the design inference.
She has objected that specified complexity and the design inference do not give a method for calculating probabilities. She is correct, but the design inference was never intended to do that. It is not about how we calculate probabilities, but about the consequences of those probabilities. Liddle is complaining that the design inference isn't something that it was never intended to be.
If you want to know whether or not Darwinism is improbable, you need to look elsewhere. Liddle contends that the design inference sneaks in an elephant, the assumption that Darwinism is improbable. But the design inference makes no pretense of providing the elephant. The elephant -- see, for example, Stephen Meyer's current book Darwin's Doubt -- must be purchased separately.
Winston Ewert is Research Assistant at the Evolutionary Informatics Lab.
Image credit: Kevin H./Flickr.