Christoph Adami’s book, The Evolution of Biological Information, promises to explain “how evolution creates complexity, from viruses to brains.” The publisher, Princeton University Press, describes the book as covering:
Why information is the unifying principle that allows us to understand the evolution of complexity in nature.
But does Professor Adami actually show how evolution can explain biological information?
Whenever someone makes a claim about biological information, the first question is what definition of information is being used. Information is not a concept with only one meaning; rather, there are different definitions and models of information. For a book such as this, it is crucial to understand what definition of information is being used in order to evaluate its claims.
Shannon’s Information Theory
Adami bases his definition of information on Claude Shannon’s information theory. A key principle behind this version of information theory is that information is always something that one thing has about something else. For example, a weather forecast has information about the actual weather. Election polling has information about future election results. Records have information about historical events.
This book adopts the formulation that genomes have information about the environment. As an illustration of this idea, consider the case of sickle cell anemia. This is a blood disorder with a number of deleterious effects. However, it also confers some degree of protection against malaria. As a consequence, it is relatively common in areas where malaria has been historically prevalent. Plausibly, this is a consequence of natural selection favoring people with this disorder in those areas.
If we were to look at a genome and determine that it had the mutation for sickle cell anemia, that would tell us something about the environment from which that genome was taken. It suggests that the person from whom the genome was sampled derives their ancestry from a part of the world where malaria was rampant. The genome thus contains information about the environment of its derivation.
The Intelligent Design Perspective
Crucially, this is not the same conception of information that intelligent design proponents are utilizing when they point to biological information. When we speak of information, we are typically referring to specified complexity, observed events that are both highly improbable and adhere to an independent pattern.
To understand the difference, consider the case of a functional protein, for example, hemoglobin, which is responsible for transporting oxygen in red blood cells. Intelligent design proponents would say that the sequence coding for hemoglobin contains information because it describes a functional protein, and thus adheres to an independent pattern, as well as being highly improbable because very few sequences would produce a functional hemoglobin.
On the other hand, under the book’s definition, hemoglobin contains information only insofar as it tells us something about the environment. We might think of it as telling us that the environment has oxygen. Or perhaps it tells us that this particular sequence codes for a functional hemoglobin. It might be thought of as containing information about the best amino acid to use in that position for that environment.
But this points to a problem with the model being used. It is unclear how one would determine what information a genome actually provides about its environment. In order to apply information theory here, we would have to begin with a model of the different possible environments that were under consideration. Only then could we evaluate how much a genome tells us about the environment.
The book resolves this by subtly shifting the definitions. Instead of attempting to quantify the information that the genome has about the environment, they compute the information the environment has about the genome. Given a particular environment, how much does that narrow down which sequences we are likely to observe?
Adami treats these as the same thing. They operate as though the information that a genome has about its environment is the same as the information the environment has about the genome. This is incorrect. The particular quantity they are using, specific information, is not symmetric. The information the genome has about the environment is not necessarily the same as the information the environment has about the genome.
Consider a Driveway
To understand the distinction, consider a driveway. Whether the driveway is wet provides information about whether it rained. If the driveway is dry, it certainly has not rained. However, if the driveway is wet, that does not prove it rained. Rather, the driveway might be wet for another reason, such as a car being washed or a sprinkler being run. Nevertheless, a wet driveway does provide information suggesting that it rained.
However, we can look at this from the opposite angle. Whether it rained provides information about whether the driveway is wet. If it rained, that tells us the driveway will be wet. If it did not rain, that tells us the driveway is likely not wet.
We cannot conflate the information the rain provides about the driveway with the information the driveway provides about the rain. Rain guarantees a wet driveway, but a wet driveway does not guarantee that it rained. The former provides more information than the latter. This book misuses information theory by making the same conflation.
In information theory, information is usually symmetric. But that is because information is usually taken to refer to mutual information whereas the book is using specific information. Mutual information is the probabilistic average over specific information. The average is symmetric, but the individual specific information that the book is using is not.
Variability in Nature
In practice, the book estimates information by considering the variability observed in nature. For example, a highly variable sequence is considered to contain low information, whereas a highly conserved sequence contains high information. This is a valid way to estimate the information the environment provides about the genome, but only if the environment is the only thing that determines the possible sequence. If, instead, something like historical contingency plays a role in the sequences, and it certainly does, then these calculations are in error.
Adami’s book uses a definition of information distinct from that of intelligent design proponents. That is his right. But it does mean that the book is not addressing the intelligent design argument. However, even on its own terms, the book fails to follow its own definition of information correctly. It conflates the information a genome has about its environment with the information that the environment has about the genome. Even there, the calculation assumes that only the environment matters for the observed sequences, and this will not be accurate.