12 November 2014

Is Model-Dependent Realism "self-refuting"?

Recent comments have prompted me to revisit an old post from semi-regular commenter Steven Jake, from his own blog The Christian Agnostic — which I have to say, despite our constant state of disagreement, I generally find to be one of the more thought-provoking religious blogs I've read.

A couple of posts back, I described myself as a non-eliminative physicalist — that I'm a physicalist because I do not believe in mystic or supernatural forces, and I believe that all phenomena in the known universe, including our consciousness, ultimately terminate in physical forms. But I'm non-eliminative because I believe that there are different levels of description of reality — that different models of reality with different semantic and theoretical frameworks can overlap, and can be equally 'real'.

This came from my relatively recent re-reading of George Lakoff's Philosophy in the Flesh (it's really long and very technical in parts, so you definitely pick up lots of stuff on a second or third pass). Lakoff draws some examples from neuroscience (he is a cognitive linguist) to illustrate his point:
We have seen that reality and truth occur relative to our understanding at many levels and from many perspectives. This is inconsistent with the classical eliminativist program in the philosophy of science, which asserts that the only realities and the only truths are at the "lowest level," here the neural level, that is, the level of neurochemistry and cellular physiology.
Virtually no neuroscientists hold this position, as can be seen by the ubiquity of the Neural Computation metaphor. Strictly speaking, "neural circuits" with their neural computational numerical algorithms are not a direct part of neurochemistry and cellular physiology, which talks about such things as ion channels, neurotransmitters, and cell permeability. But the Neural Computation metaphor, which defines the field of computational neuroscience (linking the middle level to the bottom level in the common paradigm), is absolutely necessary to an adequate understanding of how the brain and body function. No serious neuroscience could "eliminate" these higher, metaphorically constituted levels of scientific understanding at which computations using numbers are taken as real.
The same is true of the models of linguistic and cognitive behavior constructed by cognitive linguists and other cognitive scientists, for example, structures like conceptual systems and theoretical constructs such as basic-level categories, conceptual metaphors, image schemas, and prototypes. When there is sufficient convergent evidence, such theoretical constructs are taken as "real." Since we are not, and could not be, aware of them, they are postulated as part of the cognitive unconscious. What we call "the cognitive unconscious" is the totality of those theoretical cognitive mechanisms above the neural level that we have sufficient evidence for, but that we do not have conscious access to. Like each of the cognitive mechanisms that constitute it, the cognitive unconscious as a whole, as a general phenomenon, is taken to be real.

The term model-dependent realism didn't come around until many years later when Stephen Hawking wrote The Grand Design, but I think Hawking and Lakoff would find themselves in much agreement. MDR is a non-eliminative physicalist position, and squares very well with Lakoff's embodied realism. Take some of Hawking's quotes on MDR:
[Model-dependent realism] is based on the idea that our brains interpret the input from our sensory organs by making a model of the world. When such a model is successful at explaining events, we tend to attribute to it, and to the elements and concepts that constitute it, the quality of reality or absolute truth.
According to model-dependent realism, it is pointless to ask whether a model is real, only whether it agrees with observation. If there are two models that both agree with observation ... then one cannot say that one is more real than another. One can use whichever model is more convenient in the situation under consideration. 
[...] our brains interpret the input from our sensory organs by making a model of the outside world. We form mental concepts of our home, trees, other people, the electricity that flows from wall sockets, atoms, molecules, and other universes. These mental concepts are the only reality we can know. There is no model-independent test of reality.

There's clearly a great deal of commonality between Lakoff and Hawking's theses. Personally, I've found these types of epistemological frameworks exceptionally reasonable and parsimonious. It seems to me that the only way one could deny these models is to claim that you have a privileged epistemological access to reality not afforded to the rest of humanity, that the very idea of modeling and testing reality is misguided — and the entirety of scientific truth along with it. 


Steven Jake, however, has a much less charitable view. Just a few quotes from his post on the subject:
Frankly, I get quite tired of having this idea thrown around when it is so blatantly ridiculous, and thus I felt the need to demonstrate such.
It should be obvious upon first glance how extreme and ridiculous MDR is.
This is pure nonsense.
To illustrate the depravity of such a position....
Yeah…this is the intelligence blooming from the mind of Hawking, and this is why Einstein said, “the man of science is a poor philosopher”. 
I personally find it ridiculous, and therefore, not pragmatic, to hold to MDR

Let's cut through the rather thick layer of dismissive snark in this harangue and see what constructive criticisms Steven really has to offer:
[MDR] is an amalgam of previously known philosophical positions such as pragmatism, constructive empiricism, and some type of idealism—all, at least to me, untenable philosophies in themselves.
Not quite. MDR, and with it Lakoff's embodied realism (which I'm subsuming under MDR for the sake of this discussion), certainly take several important points of view from a variety of older analytic philosophies. But they also summarily reject many of the tenets these philosophies espouse, and it's important to recognize those distinctions. For his part, Lakoff's work is much more focused on dissecting various viewpoints of analytic and postmodern philosophy, and articulating why in his view they are inconsistent with empirical evidence for the cognitive structure of the embodied mind. Hawking, by contrast, doesn't spend too much time musing over philosophical positions since his book is primarily about speculative physics. I think then that Lakoff's work is probably a better choice for someone interested in contrasting an MDR-like epistemology with classical philosophy, though I don't think that excuses hasty dismissal of Hawking's proposition.

Steven continues,
First, let it be understood that MDR does not claim that if two theories can both accurately describe or predict the same observations, then we cannot, at the moment, determine which theory actually conforms to objective reality. No, MDR is claiming that neither theory conforms to reality more than the other—that is, neither theory is more real than the other.
MDR does not claim that models "conform to reality" at all; it summarily rejects the idea of an absolute reality to which we have unfettered access — this means we cannot, in principle, know whether a model "conforms to reality". Rather, it claims that our very concept of what reality is is contingent upon our ability to construct models and test them against observation.We assign the term "real" to concepts that allow us to successfully model and predict the world around us. This is quite similar to what Lakoff states (emphasis mine):
In an eliminative physicalist theory, explanation would flow in one direction only, bottom to top, with only the neurobiology taken as real and the other levels taken as epiphenomena.
This is not the case in the [Neural Theory of Language] paradigm. The paradigm is physicalist in that it does not claim that any mystical nonphysical entities such as soul, or spirit, or a disembodied Cartesian mind exist. Ultimately, the brain is all neurochemistry and neurophysiology. But it is noneliminative in two ways. First, each level is taken as real, as having a theoretical ontology necessary to explain phenomena. Second, explanation and motivation flow in both directions. To explain how the neurochemistry and neurophysiology function in networks of neurons, we need a theoretical level of neural computation. Explanation of what the physical neurons are doing flows from the middle level to the bottom level. This is a noneliminative, top-to-bottom form of explanation.
Steven has fundamentally misunderstood what MDR means in saying that no model can be said to be more 'real' than any other; it is simply saying that different 'frames of reference', such as the neural and cognitive models of the mind, overlap and converge to form our picture of reality, even though they may in some ways be semantically or theoretically incompatible (that is, no one frame of reference can fully explain all phenomena). Our confidence in what we call 'real' increases as evidence from those multiple frames of reference forms a composite picture from compatible yet distinct models.

Lakoff's neural examples are great, and an evolutionary one is great as well: our confidence in the reality of evolution is based on convergent evidence from geology, paleontology, molecular biology, and genetics. These fields use disparate methodologies with distinct semantic and theoretical frameworks, but create an overlapping and compatible picture which — because it is highly improbable that so many independent models could all be wrong — increases our confidence in the reality of evolution. Yet you cannot talk about paleontology using the semantic and theoretical framework of molecular genetics; each model requires its own unique paradigms, and we can coherently describe them both as 'real' because they both possess a theoretical ontology necessary to explain observed phenomena.

Steven continues, contrasting MDR with some well-known philosophical positions:
The realist states that an objective reality exists independent of observers, while the solipsist states that only his mind exists. Surely these models are mutually exclusive, and either one or the other has to be predicated of reality—that is, either only I exist, or a reality exists which I am a part of; there is no middle ground here. However, on MDR we cannot say that one is true while the other is false. Rather, neither is true.
MDR would say that both classical realism and solipsism (specifically, ontological solipsism) make fundamentally untenable assumptions. We do not have unfettered access to an ultimate absolute reality, and we have ample reason to assume, based on evidence arrived at through induction, that a reality external to our minds does in fact exist. MDR does not summarily declare either position false, as Steven asserts; rather, neither can be said to be true or false.

Steven then claims that MDR leads to the "rejection of itself" through pragmatism:
Remember that MDR asserts that if two models of reality are equally on par at describing our observations, then we can use whichever model we find most valuable or convenient. This is pure pragmatism—and MDR, just like any philosophy founded on pragmatism, runs into problems because of this. First, if one can adopt any model based on its utility, then the validity of models is subject to the whims of individuals. For what’s pragmatic for me will not necessarily be pragmatic for you.
Steven has begun here with a bit of a straw-man argument. Let's look again at what Hawking said:
If there are two models that both agree with observation ... then one cannot say that one is more real than another. One can use whichever model is more convenient in the situation under consideration.
Now pragmatism in philosophy is, according to the IEP, "a philosophical movement that includes those who claim that an ideology or proposition is true if it works satisfactorily, that the meaning of a proposition is to be found in the practical consequences of accepting it, and that unpractical ideas are to be rejected." What Hawking is talking about here has absolutely nothing to do with philosophical pragmatism.

The first clue is that Hawking describes a scenario in which two or more models as agree equally with observation, and in doing so has set up an important condition that prevents us from having a choose-your-model free for all. Lakoff's neural paradigm provides a fine example. The neurobiological, neurocomputational, and cognitive models of the mind all agree strongly with observation. But it's impractical for the neurobiologist to use a cognitive model of the mind, because it requires a completely different semantic and theoretical structure. For a physicist, it's entirely possible (in principle, at least) to calculate the trajectory of a baseball pitch using quantum mechanics. But given the extraordinary amount of data that would be required, it's much more practical to simply use Newton's laws of motion. Both quantum mechanics and Newton's laws agree strongly with observation, but physicists can take pragmatic concerns into account when deciding which frame of reference to use. That measured, contextual distinction is what Hawking is describing — not the ridiculous position Steven has erroneously attributed to him.

Mistakenly thinking he's scored a decisive rebuttal, Steven continues:
Pragmatism is a theory of truth which explicitly rejects the correspondence theory of truth—i.e., a statement is true if it corresponds to reality. But, the only way to do this is to make an objective claim about reality—that is, truth as such is illusory.
Steven's already wrong in claiming that MDR is a pragmatist epistemology; but now he's doing something that would make most academic philosophers cringe in claiming that it's in principle impossible to reject the correspondence theory of truth without making an objective claim about reality itself. The correspondence theory of truth, however, is not the golden child of philosophy; philosophers have long posited a number of objections to it.  The correspondence theory is epistemic, not ontological, and so is its rejection. And where the correspondence theory most finds itself in trouble is precisely in the notion that we have unfettered access to an absolute reality — that the concepts of the mind correspond directly with the world. Lakoff has identified three gaps in analytic philosophy that the correspondence theory has to overcome:
Gap 1: The gap between the natural language and the symbols in a "formal language" that are used to represent aspects of the natural language.
Gap 2: The gap between the symbols of the formal language and the sets of arbitrary abstract entities in the set-theoretical model of the language.
Gap 3: The gap between the set-theoretical models of the world and the world itself.
This is beyond the scope of this post, but suffice to say that 1) a great many philosophers have objected to the correspondence theory for a variety of reasons, and 2) MDR has nothing to do with philosophical pragmatism.

Steven wraps up,
Advocates of MDR seem to make their case solely on the basis of how our sensory organs take in datum and relay it to our brain. This seems, to them, to demonstrate that only the appearance of reality is accessible to us, as opposed to reality in itself. But, notice that this is once again an objective claim regarding what really goes on when we abstract concepts from the observed. More importantly there are many epistemologies out there that take such cognitive facts into account while still proclaiming that reality in itself can be known. 
Steven's gotten all the way to the end of his post without grasping the most important point: MDR renders meaningless the distinction between "reality accessible to us" and "reality in itself". Of course most philosophers and scientists (including Hawking) operate on the provisional, inductively-derived assumption that an absolute reality does exist. But we do not have an unfettered, privileged access to such a reality in which one level of explanation (or one 'frame of reference') successfully describes all phenomena. Rather, our picture of reality is a collage of convergent and overlapping models whose success and applicability is contextual and specific.

So has Steven successfully dismantled model-dependent realism? I don't believe so. In my estimation, he's only succeeding in betraying his hasty, uncritical dismissal of an idea to which he is personally incredulous. 

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.