Language, Science, and Perception

The dialogue continues! Responding to Richard Gipps's post contra aftab contra gipps contra seth

Gipps is quite right to point out that Seth (and other neuroscientists) have not been entirely clear about the sense in which they are using the terms such as "inference" and "prediction" when applied to neurological processes, how this usage departs from "ordinary" usage, and that this lack of clarity leads to confusion, inconsistency, and yes, possibly erroneous inferences. On this issue I am in agreement. While the project of bringing philosophical clarity to these neuroscientific terms is not necessarily easy, it is not impossible. Where I disagree with Gipps is that he seems to think that the entire project of explaining perception is muddled because there is no meaningful question to be answered, and that there is no meaningful way in which brains can be said to "infer" anything other than in a completely metaphorical manner. My own view is that there is a meaningful scientific question to be asked and answered, and there is a meaningful way in which brains can be said to predict or infer something that is not purely metaphorical.

#1. The truth about orbits

I had stated that when it comes to orbits, there are right and wrong answers: "the scientific model according to which sun is in orbit around the earth. This scientific model is decidedly false; the sun is not in orbit around the earth..."

Gipps on the other hand seems to think that when it comes to orbits, it is simply a matter of frame of reference: 

"when we're discussing the movements of celestial bodies, we might surely offer either the sun or the earth as our reference frame. We could of course stipulate the moon but, depending on the scientific context, that'd probably make the math very complex."

Gipps here is referring to the relativity of motion. He is indeed correct that all motion is relative to some frame of reference, but orbital relationships are not merely about motion. They are primarily about gravity. The movement of celestial bodies are governed by gravity, or the curvature of the space-time continuum per Einstein's general relativity. Yes, if we take Earth to be the frame of reference, the relative movement of the Sun is around the Earth, but there is no scientific explanation as to why Sun would move around the Earth. On the other hand, there is a perfectly good explanation as to why Earth would move around the Sun. This is because Sun is extraordinarily more massive than Earth, it has a much larger gravitational pull, and it curves the space-time continuum around it. So contra Gipps, the truth of orbital relationships is not a matter of how simple the mathematical model of relative motion is, but rather what is the explanation that governs that relationship.

To restate my original point, it appears to us that Sun is in orbit around Earth because of relative motion, but Sun is not in orbit around Earth because that's not how gravity works. It is decidedly false because it contradicts our deepest, most reliable, most empirically supported theories of gravitation.

#2. Language and Science

Just as Gipps seems to think that asking "does the earth orbit the sun, or does the sun orbit the earth?" just isn't a good question and we don't really know what the question means, Gipps seems to think that "how do we perceive?" is not a meaningful question either. I think it is perfectly meaningful to ask "does the earth orbit the sun, or does the sun orbit the earth?", as I've explained above, and similarly I think it is perfectly meaningful to ask how do we perceive. It seems to me that just as Gipps doesn't quite appreciate that we can't understand how orbits work without referring to gravity, we can't quite understand how perception works without referring to... whatever the best explanatory theory for perception will turn out to be.

#3. Acceleration and Perception

Dr Gipps uses the example of a car accelerating: "If I want to know how, say, a Porsche can accelerate so quickly - the acceleration admittedly being a property of the whole car - I'll naturally be satisfied with an answer in purely mechanical terms, one which tells me about the functioning of the carburettor etc. Sure, I'll want to know too how the carburettor is connected to the throttle and the fuel supply and the engine and thereby to the wheels - but there seems to me no requirement for an extra kind of story, told at some other level of explanation, wherein I relate all of this to the accelerating car."

I don't find this very satisfying, because acceleration is a very different beast than perception. 

Acceleration is not merely a property of the whole car. We can talk about parts of the car -- such as the wheels -- accelerating as meaningfully as we can talk about the whole car. Not only that, acceleration is not something that emerges only at a certain level or organization. Particles can accelerate. Billiard balls can accelerate. Cars can accelerate. Comets can accelerate. 

There is also an explanation that connect the carburetor and the acceleration of the car; the carburetor provides energy needed for acceleration by the process of combustion. Energy is needed for acceleration, because acceleration is the rate of change of velocity per unit of time, and velocity cannot change without a change in energy.

The case of perception is different in crucial ways. The DNA doesn't perceive. An individual neuron doesn't perceive. But a person or organism perceives. How does that happen? We don't quite know. Unlike energy from fuel linking carburetor and acceleration, we don't have an obvious explanation that links voltage changes in the nerve membranes to an organism perceiving. A carburetor doesn't need to make anything analogous to inferences, but maybe, just maybe, the brain does.

Gipps: "Now I'm not sure what it means to talk of a brain being 'confined' (perhaps 'safely contained'?) in a skull, and of it only having 'access' to this or that sensory signal."

It may be helpful to think of this problem in terms of 'information'. Now Gipps may say here that he is not clear at all what 'information' means outside the context of knowledge in a human mind. To offer some clarity, I am using the term in the same sort of way a physicist talks about 'information' being lost in a black hole (cf. Black hole information paradox). What information means in this physical context is that the state of a system at one point in time has a discernible relationship with the state of a system at any other time (e.g. you can use an equation to calculate the state of a system at one point given the state of the system at another point). In the case of perception, let's say I see a tree in front of me, and then I copy the shape of the tree on a piece of paper. We can think of this in terms of flow of 'information' -- there is a relationship between the physical state of the tree, the physical state of my brain, and the physical state of the piece of paper. The relationship between the physical state of the tree and the relevant physical state of my brain is mediated by the sensory nerves. The 'information' that reaches the brain reaches only in the form of voltage changes in the membranes of the sensory nerves. But what is the nature of the relationship between the state of the brain and the state of the nerves? And what is the nature of the relationship between voltage changes in the nerve membranes and the actual tree? We can think of neuroscientists such as Seth trying to provide an answer to this question. 

To make matters complicated, while the brain contains 'information' in the physical sense described above, my mind contains information in the ordinary sense (I acquire information about the tree by perceiving the tree). What is the relationship between the information in my mind (ordinary language sense) and information in my brain (physical sense)? These two sorts of information cannot be entirely disconnected! If there was no relationship between the two, my knowledge of the tree would be miraculous, with no natural explanation.  

No such mysteries about 'information' arise in the case of an accelerating car. There is nothing unexplained about the flow of energy.

#4. What does it mean for a system to enact a mathematical model?

I have something different in mind than the models of how celestial bodies move or how a pancreas operates. These would be pure metaphors. We can talk as if pancreas is making a prediction, and we may usefully speak so for practical purposes, but there is nothing like prediction actually happening (as far as we know). What I have in mind is a system doing something that is analogous to humans predicting. Consider the technology of "predictive text." As I am typing a message on gmail, their predictive text feature is trying to "predict" ("anticipate" "guess" "project") how I will finish the sentence. Unlike pancreas, there is something happening here: a complicated algorithm that has gone through thousands and thousands of my prior emails has detected patterns of how I use words, and uses those patterns to anticipate what I will type next. Is this literally prediction in the ordinary sense? Maybe not. But it is analogous enough, it is similar enough, that we can extend the usage and it will remain meaningful. 

Prediction as metaphor: there is nothing like prediction (ordinary sense) happening, but we are talking as if there were

Prediction as analogy: there is something like prediction (ordinary sense) happening; it is not identical to prediction (ordinary sense), but it is similar enough that the use of the term can be meaningfully extended.

If there is something in the brain that approximates Bayesian inference, then something like prediction and inference can be said to be taking place.

#5. Linguistic innovation

Gipps: "Aftab effectively claims that I was saying that unless Helmholtz, Seth, whoever, offers an explicit definition of a term used in a new sense, that new use must be considered nonsensical or muddled. But this really wasn't at all what I was saying"

I accept that I was being unfair in this regard, and Gipps has clarified this quite well.

#6. On Use

Gipps: "whether a community's deployment of an expression is meaningful is not something guaranteed simply by their using it. And whether it's meaningful will instead amount to whether it's use can be elucidated, whether it avoids oscillating unstably between different senses encouraging the making of illicit inferences, whether clear negations of propositions deploying the term can be formulated, and sometimes, yes, even whether it can be clearly defined."

That seems quite reasonable to me. Gipps may disagree, but I do think that describing perception as a process of inference, eg, is something that can be elucidated and something that can be used consistently without confusion, even if actual neuroscientists have done a poor job at doing so! I would also say that use vs misuse may not always be clear, and when it comes to extension of ordinary use to analogous cases, judgments of whether that constitutes use or misuse may differ, even among members of a linguistic community. When it comes to whether brains can meaningfully infer or predict, who belongs is the relevant linguistic community? And how would we determine what the linguistic community deems to be correct in cases where a new extension of an old term is being considered? How are disagreements within a linguistic community to be resolved? Can an entire linguistic community be wrong about meaningfulness of a new usage? I can't say I have the answers to these questions, but I sure as hell don't think there are always easy answers here!