Analogies, Metaphors, and Inference in Science

Mike Gene, the pseudonymous founder of Telic Thoughts and the author of The Design Matrix: A Consilience of Clues has written the following about metaphors and their application to the detection of intelligent design:

Metaphors such as “fear”, “cost”, “abhor” and “angry”, commonly share the projection of consciousness onto the world. Metaphors such as these represent the human tendency to view the world through anthropomorphic glasses. However, the metaphors employed by molecular biologists are not of this type.
...
Metaphors typically break down when we begin to take them literally.
...
[but] The design terminology that is used in the language of molecular biology does not break down when interpreted literally
...
[T]here is a basic and literal truth to the use of design terminology in molecular biology–these technological concepts are just too useful. Metaphors are certainly useful when explaining concepts to other human beings, yet the design terminology often goes beyond pedagogy–it provides true insight into the molecular and cellular processes. An understanding of our own designed artifacts, along with the principles required to make them, can guide the practice of molecular biology.

Metaphors are a special case of analogies (the other being similes). I have written extensively on the subject of the use of analogies and metaphors in science, and especially in the evolution/design debate here, here, and here. The fundamental question in this ongoing debate is, how do we know an analogy really exists? For example, do we have any objective way to determine if one rock is analogous with another? Or whether an anatomical feature (or a protein/substrate binding site) is analogous to another?

As in the case of agency detection, we think we can do this very easily (just as we can easily identify what looks like design), but I would argue that this is because both "finding" analogies and "finding" design and purpose are capabilities of the human mind (and the nervous system that supports it) that have conferred enormous adaptive value on our ancestors. As in the case of our hypothesized "innate agency/design/purpose detector" (which first becomes active in very early infancy), our "analogy detector" also appears to become active at a very early age, and operates entirely in the background. That is to say, we are almost totally unaware of its operation, and concentrate only on its output.

Our ability to detect (and construct) analogies is, IMO, the core of our intelligence, as demonstrated by the fact that identifying analogies has been traditionally used as one of the most sensitive gauges of general intelligence in intelligence tests (such as the Miller Analogies Test).

In the context of Mike Gene's ideas about metaphors, doing engineering (and especially mathematics) is essentially the construction of highly compact analogies, in which numerical (and sometimes physical) relationships are expressed in the form of abstract symbols. A blueprint is simply a metaphor for a building, just as a chemical formula is a metaphor for the product of the chemical reaction, and a recipe for a cake is a metaphor for the cake. Indeed, many mathematical relationships (especially in the natural sciences) are expressed as equations, which are quite literally metaphors expressed in symbolic form. For example, Newton's equation for force:

F = ma

is a metaphor linking the concept of force with the concepts of mass and acceleration.

In molecular biology we encounter once again the concept of metaphors, for what is the genome of an organism but a highly abstract metaphor for the fully embodied and operating organism itself? I agree with those (and I expect Mike Gene would number himself among them) who assert that the encoding of "life" into a string of nucleotides is indeed the crucial difference between biological "metaphors" and physical "direct necessities". Gravity isn't "encoded" in anything, but proteins are, and so are cells, tissues, organs, organ systems, organisms, and (at least at some level) their behaviors.

So, is there a way to verify if an analogy or metaphor is "real"? In the case of some analogies in biological systems we have an independent double-check on our identification of analogies. This is based on the evolutionary concept of homology, or derivation from a common ancestor. If two structures on two different organisms (say a small bone of the jaw of a reptile and the even smaller bone in the middle ear of a mammal) appear to be analogous (on the basis of size, location, relationship to other bones, etc.) there are at least two different, though related, methods of verifying that these structures are indeed analogous (and not just accidentally similar).

One way is by means of comparative paleoanatomy, in which a series of fossils of known age are compared to determine if there is a connection between the evolutionary pathways of derivation of the structures. If such a pathway can be empirically shown to exist, this would be strong evidence for both the analogous and homologous nature of the objects.

Alternatively one could compare the nucleotide sequences that code for the structures to determine if they are sufficiently similar to warrant a conclusion of homologous derivation. In both cases, evidence for homology, combined with our intuitive "identification" of analogous structure and/or function, both point to the same conclusion: that the two structures are both analogous and homologous.

This is why structures that appear to be analogous, but for which there is no convincing evidence of homology (as in the wings of birds and insects) can present a serious problem to evolutionary biologists, and especially those engaged in biological classification. Such apparent similarities (technically called homoplasies) can either be the result of "true" (i.e. partial) analogy at the functional (and/or structural) level (and therefore assumed to be the result of convergent evolution) or they can be completely accidental. Simple inspection is often insufficient to separate these two hypotheses, and lacking either fossil or genomic evidence, conclusions about the validity of such analogies can be extremely difficult to draw. However, if there is fossil and/or genomic evidence and it points away from homology (i.e. descent from a common ancestor), then the structures can be considered to be analogous, but not homologous.

One of the dangers in invoking analogies and metaphors is overusing the concept of analogy to mean almost anything. Indeed, it is essential in discussions such as these that we be as precise as possible about our definitions, as imprecision can only lead to confusion (at best) and unsupportable conclusions (at worst).

Perhaps the most essential distinction to be made in this regard is between "analogies of description" (which could also be called "semantic analogies") and "analogies of function/structure" (which could also be called "natural analogies"). The former (i.e. "semantic analogies") are merely artifacts of the structure of human cognition and language, as happens whenever we describe an analogy that we have perceived.

By contrast, the latter (i.e. "natural analogies") are the actual similarities in function/structure that we are describing (i.e. that resulted in our identification and description in the first place). As in the Zen koan about the roshi and the novice in the moonlit garden, much of the confusion about which of the two types of analogies we are discussing seems to stem from confusion between the moon that illuminates the garden and the finger pointing at the moon.

In the brief example from Mike Gene's The Design Matrix posted at the head of this thread, the implication is that the analogies we perceive between biological systems and those engineered by humans are "natural analogies"; that is, they are real analogies, and not simply a form of linguistic convenience. However, there is nothing about the finding of an analogy that necessarily verifies that the analogy is "natural" (i.e. "real"), as opposed to "semantic" (i.e. "imaginary"). This would be the case even if one found repeated analogies between complex systems engineered by humans and biological systems that evolved by natural selection. To verify that an analogy is "natural" requires an independent source of validation for the assertion that the analogy is "real" and not merely "semantic". At this stage in my reasoning about this subject I am not at all sure how one would go about this.

However, one thing I am sure of is that simply asserting over and over again that one has perceived an analogy, and that this is all that is necessary to validate the analogy, is not sufficient. Even I am but mad north-north-west: when the wind is southerly I know a hawk from a handsaw.

************************************************

As always, comments, criticisms, and suggestions are warmly welcomed!

--Allen

"Did Anyone Observe the Origin of Eukaryotes?"

AUTHOR: Allen MacNeill

SOURCE: Original essay

COMMENTARY: That's up to you...

A creationist asked the following pretty typical question on another blog:

"Have prokaryotes ever been observed become eukaryotes via endosymbiosis?" [sic]

If by "observed", one means directly observed, then of course the answer is "no". As far as we can tell, this probably happened more than a billion years ago. But if only things that have been directly observed are valid, then virtually all of science, if not almost all human intellectual endeavors, are invalid and pointless.

Did anyone alive today "observe" the decline and fall of the Roman empire? Of course not. So, how do we know it happened? We read about it, or were told about it. We might also have directly observed some ruins in Rome or elsewhere in Europe or Asia Minor, and made some inferences about where they came from and how old they are.



But if direct observation is necessary to validate an assertion, then each of us is trapped in a tiny world whose borders are the limits of our own unaided perceptual apparatus. Not even most forms of logic would survive such an absurd and self-destructive limitation.

However, if one allows for indirect observation and logical inference, then the answer is "yes". There are multiple sources of empirical evidence for the assertion that eukaryotic cells arose as the result of the serial endosymbiosis of several prokaryotic ancestors. You can read a summary of this evidence here (scroll down; it's toward the end of the article).

Furthermore, this inference is made using the most reliable (i.e. "strongest") form of logical inference known to us: consilience. There are multiple, independently discovered and derived lines of empirical evidence pointing to the serially endosymbiotic origin of eukaryotic cells. That is, the evidence for the serial endosymbiosis theory is based on consilience, which is much more reliable that induction, deduction, or even abduction alone.

As in any case having to do with a very complex universe, there are "gaps" in our current model of the serial endosymbiotic origin of eukaryotes. There is also empirical evidence that is not entirely consistent with the model as it now stands. However, as more and more empirical evidence has been discovered, the vast majority of it has supported Lynn Margulis' original theory.

So, which method of validation shall we choose? Shall we voluntarily blind ourselves to the only kind of evidence that can validate things that have happened outside of our immediate perceptual environment, or accept what virtually all thinking people accept – that we must, almost everywhere and at almost all times, accept the validity of empirical evidence that we have not ourselves immediately obtained?

As always, comments, criticisms, and suggestions are warmly welcomed!

--Allen

Inference and the Boundaries of Science

AUTHOR: Hannah Maxson

SOURCE: Evolution and Design

COMMENTARY: Allen MacNeill

The now-notorious Cornell "evolution and design seminar" met for the first time last night, and in my opinion our first meeting was a rousing success. As I had hoped, the participants began to make their opinions and positions known (despite my blathering), and a good time was had by all. We're getting ready to analyze Richard Dawkins' arguments in The Blind Watchmaker, discussion of which will be facilitated by Will Provine (one of our faculty participants).For a brief taste of how things went last night, you should check out the course blog. Here's a sample:

Hannah Maxson (founder of the Cornell IDEA Club) wrote:

In class last night Allen went over inference and his views of the boundaries of science. He gave us the example of an individual coming upon the remains of what appeared to have been a house fire in the past. Without any prior knowledge of the event or eyewitnesses to question, one might infer any of three things (see diagram, above):

1) accidental house fire
2) arson: purposeful house fire
3) no fire at all; setup job (for film, etc.)

A tentative explanatory filter with which to distinguish between those three causes. But he suggested there is a problem from the very beginning. The first question– was this a real fire, or a setup job? can never be definitely answered. Considering a very powerful film crew, for instance, the setup would look almost like a real fire. Extrapolating slightly, given an omnipotent “designer”, could the scene not be exactly the same as what one would expect from a housefire?

Because there is no way of giving a definite answer based on empirical evidence– to which we, as scientists, are limited– we must throw out that whole node on our explanatory filter. Everything above the dotted line, at least, is outside our realm of knowledge.

I had a quarrel with much of this reasoning, though to begin with I ought to make a strong disclaimer that I’m not at all interested in defending “setup jobs”– I think they are highly uninteresting, for one thing, and not worth spending time in. But a “right” or at least convenient answer doesn’t make the logic that goes into an argument sound.

First, can we throw a question out of the realm of science because we will never be able to get a definite answer? Scarcely anything in science will ever be proved or disproved. In general, we don’t look for certain proofs, but simply for empirical evidence that might favor one or the other, so that we can make an inference to the best explanation. If the evidence is not clear, we often make choices based on conventions, such as parsimony.

If we cannot throw it out for lack of a definite answer, can we at least throw out that node for lack of empirical evidence either way? It is true that if the scene was designed (omnipotently) so that there was absolutely no evidence there had been no real fire, science could do nothing with the question. But we cannot assume a priori that all “setup jobs” have no emperical evidence available; there are a great many other possibilities besides an omnipotent designer who works to make things exactly the same. Consider, for example Einstein’s view: “Nature hides her secrets because of her essential loftiness, but not by means of ruse.”; or in another remark: “God is slick, but he ain’t mean.”

So while we can do away with a “absolutely perfect imitation” possibility as an option that could never have any emperical grounds, that is not justification for demarcating the entire first node out of our field of inquiry. In any research project you learn quickly that things are not always as they first appear. What seems on first analysis to be the remains of a fire may turn out on further investigation to hold evidence of a set-up job. What appears to have been designed may in fact be the product of chance and necessity, and what we are used to thinking of as the products of unguided evolution may contain evidence of purposeful design.

Refusing to consider questions is never good practice; we may reject explanations for lack of warrant, but ought never reject the investigation a priori.

To which I replied:

Thanks, Hannah, for the diagram (it’s clearer than mine was last night) and for your analysis, above. However, I still stand by my position that, given a sufficiently powerful “designer,” a house fire (or anything else) can be simulated to such a degree (as Warren [Warren Allman, director of the Paleontological Research Institute and Museum of the Earth here in Ithaca] said, “right down to the subatomic particles) that there would be absolutely no way to distinguish between such a creation ex nihilo and the real thing.

That is, no amount of empirical evidence could make it possible to get past the first branch point in the explanatory filter in the diagram. Indeed, every piece of empirical evidence one could add would simply amplify one’s assertion of the hypothesis of the Designer’s omnipotence (”Amazing, S/He/It can f/make things right down to the quarks!”). For this reason, rather than agonize over our inability to get past the first branch point in the filter via empirical means, we simply agree to skip that step and move down to the second branch point.

I believe that this “agreement” is something with which most ID supporters would concur, as it gets us out of an empirically insoluble dilemma, and moves us along to the question of accident vs design. Darwin did essentially the same thing in the Origin of Species, by bringing in “the Creator” only at the very end, and by relegating Her/Him/It to setting the whole system in motion in the beginning. Having spent many years reading Darwin’s personal writings (correspondence mostly, but also some of the expurgated sections of his autobiography), it appears to me that Darwin became a Deist about the time he wrote the Origin (or in the process of doing so, which took two decades), but then slowly realized that Deism is essentially equivalent to agnosticism/atheism, as the Deity of Deism plays no part in the actual universe at all, beyond setting up the natural laws that govern it. I find myself in the same situation: assuming that the Deity of Deism exists gets one absolutely nowhere at all in science, and so (like most other scientists), I simply don’t go there anymore.

And now I would go further; while it is a good idea to "not reject explanations for lack of warrant, bu never reject the investigation a priori", the point I was trying to make in my reply was that if one can't get by the first branch point in the "explanatory filter" I posited during the discussion, then we can't really do science at all. Furthermore, agreeing that the remains of what looks like a house fire could have been created ex nihilo by a sufficiently powerful entity gets us absolutely nowhere in terms of explaining the origin of the wreckage. In fact, it forestalls the possibility of any kind of empirically verifiable (or falsifiable) hypothesis, and is therefore a "science stopper" of the first order.

--Allen

Incommensurate Worldviews

AUTHOR: Allen MacNeill

SOURCE: Original essay

COMMENTARY: That's up to you...

I am beginning to understand more about the differences between the physical sciences (such as astronomy, chemistry, and physics) and the biological sciences, and why the worldview of a physical scientist with a strongly mathematical predilection is apparently so different from mine and that of most other biologists (at least, of those biologists of whom I have personal and/or reputable knowledge). Furthermore, it seems to me that these differences are central to the apparent inability of non-biologists to fully comprehend the "darwinian" worldview upon which much of biology (and all of evolutionary theory) has been constructed (and vice versa, of course).

To me, these appear to be the basic differences that inform our worldviews:

1) CONTINGENCY: The biological sciences (i.e. anatomy & physiology, parts of biochemistry, botany, development & embryology, ecology, ethology, evolution, genetics, marine biology, neurobiology, and the allied subdisciplines), like the "earth sciences" (i.e. atmospheric sciences, geology, etc.) are both contingent and historical. That is, they cannot be derived from "first principles" in the way that algebra, calculus, geometry (both euclidean and non-euclidean), probability, symbolic logic, topology, trigonometry, and other "non-empirical" sciences can be. As both Ernst Mayr and Karl Popper have pointed out, historical contingency is inextricably intertwined with biological causation, in a way that it is not in mathematics and the physical sciences. This would appear to be true, by the way, for both "darwinist" and ID models of biological evolution and the fields derived from them. Indeed, even the Judeo-Christian-Muslim worldview is contingent and historical, in ways antithetical to both mathematics and pre-"big bang" cosmological physics.

2) UNIVERSALITY: The biological sciences are also not "universal" in the way that chemistry and physics are. We assume that the processes described by physical "laws" are universal and ahistorical. that is, we assume that they are the same regardless of where, when, and by whom they are investigated. Furthermore, it is tacitly assumed by physical scientists that the "laws" they discover apply everywhere and everywhen, without empirical verification that this is, in fact, the case. It seems to me that this assumption is reinforced by the mathematical precision with which physical processes can be analyzed and described.

By contrast, the entities and processes studied by biologists are necessarily "messy" and often "non-quantifiable," in the sense that they cannot be entirely reduced to purely mathematical abstractions. The great beauty and elegance of Newton's physics and Pauling's chemistry are that the objects and processes they describe can be so reduced, and when they are, they reveal an underlying mathematical regularity, a regularity so precise and so elegant that one is tempted to believe that the mathematical formalism is what is "real" and the physical entities and processes that they describe are, at best, somewhat imperfect expressions of the underlying perfect regularities.

To me, however, what has always been appealing about biology is its very "messiness." As the so-called Law of Experimental Psychology states "Under carefully controlled conditions, the organism does whatever it damn well pleases." Biological entities and processes are not quantifiable in the same way that physical ones are. This is probably due to the immensely greater complexity of biological entities and processes, in which causal mechanisms are tangled and often auto-catalytic.

3) STOCHASTICITY: The biological sciences are irreducibly statistical/stochastic, in ways that neither the physical nor mathematical sciences generally are (although they are becoming moreso as they intrude deeper into biology). R. A. Fisher was not only the premier mathematical modeler of evolution, he was also the founder of modern statistical biometry. This is no accident: both field and laboratory biology (but not 19th century natural history) depend almost completely on statistical analysis. Again, this is probably because the underlying causes for biological processes are so multifarious and intertwined.

Physicists, chemists, and astronomers can accept hypotheses at confidence levels that biologists can never aspire to. Indeed, until recently the whole idea of "confidence levels" was generally outside the vocabulary of the physical sciences. When you repeatedly drop a rock and measure its acceleration, the measurements you get are so precise and fit so well with Newton's descriptive formalism that the idea that one would necessarily need to statistically verify that they do not depart significantly from predictions derived from that formalism seems superfluous. Slight deviations from the predicted behavior of non-living falling objects are considered to be just that: deviations (and most likely the result of observer error, rather than actual deviant causation). Rarely does any physical scientist look at such deviations as indicative of some new, perhaps deeper formalism (but consider, of course, Einstein's explanation of the precession of the orbit of Mercury, which did not fit Newton's predictions).

4) FORMALIZATION: There are many processes in biology, and especially in organismal (i.e. "skin out" biology) that are so resistant to quantification or mathematical formalization that there is the nagging suspicion that they cannot in principle be so quantified or formalized. It is, of course, logically impossible to "prove" a negative assertion like this - after all, our inability to produce a Seldonian "psychohistory" that perfectly formalizes and therefore predicts animal (and human) behavior could simply be the result of a deficiency in our mathematics or our ability to measure and separately analyze all causative factors.

However, my own experience as a field and laboratory biologist (I used to study field voles - Microtus pennsylvanicus - and now I study people) has instilled in me what could be called "Haldane's Suspicion:" that biology "is not only queerer than we imagine, but queerer than we can imagine." That is, given the complexity and interlocking nature of biological causation, it may be literally impossible to convert biology into a mathematically formal science like astronomy, chemistry, or physics.

But that's one of the main reasons I love biology so much. Mathematical formalisms, to me, may be elegant, but they are also sterile. The more perfect the formalism, the more boring and unproductive it seems to me. The physicists' quest for a single unifying "law of everything" is apparently very exciting to people who are enamored of mathematical formalism for its own sake. But to me, it is the very multifariousness – one could even say "cussedness" – of biological organisms and processes that makes them interesting to me. That biology may not have a single, mathematical "grand unifying theory" (yes, evolution isn't it ;-)) means to me that there will always be a place for people like me, who marvel at the individuality, peculiarity, and outright weirdness of life and living things.

5) PLATONIC VS. DARWINIAN WORLDVIEWS: It seems to me that many ID theorists come at science from what could be called a "platonic" approach. That is, a philosophical approach that assumes a priori that platonic "ideal forms" exist and are the basis for all natural forms and processes. To a person with this worldview, mathematics are the most "perfect" of the sciences, as they literally deal only with platonic ideal forms. Astronomy, chemistry, and physics are only slightly less "prefect," as the objects and processes they describe can be reduced to purely mathematical formalisms (without stochastic elements, at least at the macroscopic level), and when they are so reduced, the predictive precision of such formalisms increases, rather than decreases.

By contrast, I come at science from what could be called a "darwinian" approach. Darwin's most revolutionary (and subversive) idea was not natural selection. Indeed, the idea had already been suggested by Edward Blythe. Rather, Darwin's most "dangerous" idea was that the variations between individual organisms (and, by extension, between different biological events) were irreducibly "real." As Ernst Mayr has pointed out, this kind of "population thinking" fundamentally violates platonic idealism, and therefore represents a revolutionary break with mainstream western philosophical traditions.

I am and have always been partial to the "individualist" philosophical stance represented by darwinian variation. It informs everything I think about reality, from the idea that every individual living organism is irreducibly unique to the idea that my life (and, by extension, everybody else's) is irreducibly unique (and non-replicible). Such a philosophical position might seem to lead to a kind of radical "loneliness," and indeed there have been times when that was the case for me. But since all of us are equal in our "aloneness," it paradoxically becomes one of the things we universally share.

And so, I don't think a "darwinian worldview" applies to the physical sciences (and certainly does not apply to non-empirical sciences, such as mathematics), for the reasons I have detailed above. In particular, it seems clear to me that although it may be possible to mathematically model microevolutionary processes (as R. A. Fisher and J. B. S. Haldane first did back in the early 20th century), it is almost certainly impossible to mathematically model macroevolutionary processes. The reason for this impossibility is that macroevolutionary processes are necessarily contingent on non-repeatable (i.e. "historical") events, such as asteroid collisions, volcanic eruptions, sea level alterations, and other large-scale ecological changes, plus the occurrence (or non-occurrence) of particular (and especially major) genetic changes in evolving phylogenies. While it may be possible to model what happens after such an event (e.g. adaptive radiation), the interactions between events such as these are fundamentally unpredictable, and therefore cannot be incorporated in prospective mathematical models of macroevolutionary changes.

It's like that famous cartoon by Sidney Harris: "Then a miracle occurs..." The kinds of events that are often correlated with major macroevolutionary changes (such as mass extinctions and subsequent adaptive radiations) are like miracles, in that they are unpredictable and unrepeatable, and therefore can't be integrated into mathematical models that require monotonically changing dynamical systems (like newtonian mechanics, for example).

So, to sum up, I believe that the "darwinian worldview" applies only to those natural sciences that are both contingent and intrinsically historical, such as biology, geology, and parts of astrophysics/cosmology. Does this make such sciences less "valid" than the non-historical (i.e. physical) sciences? Not at all; given that physical laws now appear to critically depend on historical/unrepeatable events such as the "big bang," it may turn out to be the other way around. In the long run, even the physical sciences may have to be reinterpreted as depending on contingent/historical events, leaving the non-empirical sciences (mathematics and metaphysics) as the only "universal" (i.e. non-contingent/ahistorical) sciences.

To summarize it in a bullet point:

• Platonic/physical scientists describe reality with equations, whereas darwinian/biological scientists describe reality with narratives.

--Allen

P.S. Alert readers may recognize some of the hallmarks of the so-called Apollonian vs. Dionysian dichotomy in the preceding analysis. That such characteristics are recognizable in my analysis is not necessarily an accident.

P.P.S. It is also very important to keep in mind, when considering any analysis of this sort, that sweeping generalizations are always wrong ;-)