I don't think there is any question that many ID claims rely heavily on arguments from analogy. Since the analogies drawn by ID proponents do have prima facie power (even Dawkins admits this: "Biology is the study of complicated things that give the appearance of having been designed for a purpose."), the question that remains is how much force this type of scientific reasoning carries. Typically ID critics attempt to play down the power of analogical arguments. Now clearly, use of analogy in a superficial way (if it looks like a duck, walks like a duck, and quacks like a duck, it's a duck) carries little force and can lead to false conclusions. To determine if it's really a duck a more in-depth probing should be required. Perhaps it's really a machine designed to simulate a duck but underneath the feathers it has a metallic frame, motors, gears, and levers. So what form of exploration and reasoning is required to validate the conclusion? Is there really something more than the utilization of analogy that can be employed to overcome the weakness of a prima facie claim? Or is analogy really the foundational engine driving reason and scientific validation? In other words, is it analogy "all the way down" If so then the question is not whether arguments from analogy are valid and powerful, but what analogical arguments are being made and how compelling are they?
To answer this question let's look at some science. Prominent neuroscientist and philosopher Paul Churchland looks at this very thing in his Engine of Reason, the Seat of the Soul. Now before someone thinks the reference to "soul" in the title indicates that Churchland is attempting some religious apologetics, think again. To the contrary Churchland is an eliminative materialist. However, his book does offer some very interesting research on how the brain works and how reasoning occurs. His model is a connectionist model illustrating how neural trajectories and multidimensional, neuronal activation vectors drive reason. I'll try to summarize his assertions as they apply to this topic.
First Churchland claims that as we experience life we develop "prototypical representations" of things. He uses face recognition as an example.
The brain seems to represent faces with a pattern of activations in a special cortical area somewhat farther along in the visual system(the occipito-temporal region), a pattern whose elements correspond to various canonical features or abstract "dimensions" of observed faces.
The human family displays a wonderful diversity of faces, but each one strikes out in its own idiosyncratic direction from what might be call the standard, average, or prototypical human face.
This provides for the ability to recognize a face as a face even if they have varying attributes. Then as new faces are experienced the activation matrix of the prototype is fine tuned and expanded.
Another interesting and important feature of neural networks is their ability to "fill in" missing information. He calls this vector completion. I'm sure everyone has had experiences where they seem to be able to recognize something from only a small part of the data. Activation vectors have this ability.
This ability to identify familiar faces correctly despite the loss of 20 percent of the information in the input vector illustrates a marvellous property of neural networks, a property with far-reaching consequences. This property is their capacity for what is called vector completion. Since the output layer identifies the input person correctly as Jane, it must be responding to a facial coding vector at the second layer that is very close to the proper coding point for Jane. At the least, the second-layer coding vector must be closer to Jane's point than to any other person's coding point, else the output layer could not have identified her correctly.
Apparently this ability is extremely important for creative thought. In fact, Churchland thinks it is vector completion (filling in) that is the basis of inductive inference.
What we have observed here, in the phenomenon of vector completion at layers two and three, is a primitive form of inductive inference. Indeed, it may be that vector completion is the basic form that inductive inference takes in living creatures in general.
Now prototypes shouldn't be thought of as some isolated neural representations. In fact they are interconnected with other prototypes. In neural networks activation vectors share neurons and connections with other activation vectors. According to Churchland that interconnectivity offers the ability for forming novel combinations of activations to create new "concepts". It is this ability that makes scientific progress possible. When the scientist is confronted with something new, certain sets of activation vectors are fired possibly in ways not activated before. This can account for the generation of new ideas and explanations.
To illustrate this Churchland uses an example of translating the observation of wave interference in a sea wall to the problem of the double slit experiment in quantum mechanics. Now the dynamics of a sea wall is observable.
When the double slit experiment is done, if the experimenter initially thinks that light is a particle, for very thin slits the pattern on the screen is puzzling. This is because instead of two very bright images there are several bright lines with the brightest in the center and the fainter ones on either side. This would be very puzzling but for the fact that a similar pattern can be observed with a sea wall. Since this prototype would already be in place in the scientist's mind, the analogy can be applied to what is unobservable and the idea that light can behave like waves emerges. This is a perfect example of the how the brain uses analogy to make scientific discoveries. Churchland states:
The prototype's novel application to optical phenomena is systematically vindicated by the prototype-driven experimental probing. The distribution of the bright and dark bands of light vary exactly as do the high- and low-amplitude water wave sites. What began as an analogically inspired guess quickly acquires the status of a confidently held theory. If the manipulative powers that characterized the original prototype carry over successfully into its new domain of deployment, wild horses will not stay our conviction that light must be waves.
Now ID critics might claim that analogical reasoning is only the starting point, could be fallacious, and must be validated by empirical data. Surely this is true. However, data is just data. It tells us nothing unless it is interpreted. But according to Churchland's model interpretation relies on analogy as well. It would seem that science is absolutely riddled with analogical thinking "all the way down". To strip analogical reasoning of its force both in the beginning of scientific investigation and its subsequent fleshing out would seem to hamstring it entirely. It appears that fundamentally the brain is an analogical machine, comparing observed patterns with prototypes and integrating them in new ways with other activation vectors. This in turn offers new creative possibilities for explanation (abduction?). One has to wonder what prototypes Darwinists are utilizing when they claim that complex biotic systems are the result of blind, non-intentional processes.
Now, if ID was to base its claims on superficial analogy, it would rightly be rejected or at least held under high suspicion. However, as biotic systems are probed at deeper and deeper levels, the analogies to design just keep popping up. In these systems we find prototypes of complex information, manufacturing, transport systems, signal transmission, sensors, error checking, redundancy, built-in-test, modularity, sequenced assembly instructions, timed activations, feedback systems, etc. The list goes on and on. The fact that these and many other features found in biology are analogous to design as humans know it, is evident in the engineering and teleological language that is used to characterize it. If analogical thinking is "all the way down" in how science is done, how many analogies does it take for the scientific community to finally agree that "it's designed" is worth exploring?