Who Needs Evidence?

We’re done, today, with the main part of the course. Starting next time we’ll be reading a selection of essays from Ammerman. Most of them relate to the main topic. For example, next time we’ll be reading an article by G. E. Moore about the way in which I know my hand is in front of my face.

The final chapter is largely about evidence, making the point that as soon as you raise the questions, "What is the evidence by which I know that I have my hand in front of my face?," "that I live in Tucson?," "that I own a dog?," the game is already lost. One did want evidence that there were weapons of mass destruction in Iraq, but that is precisely because it was in serious doubt whether there were WMD in Iraq. There is no doubt whatever that I live in Tucson, own a dog, …. It is the misleading question, "What is your evidence?" that makes it seem plausible that there is any doubt.

So the right question to ask is, "Why would anyone think evidence might be required?" The answer, of course, in Ayer's case and in Warnock's is foundationalism. If there is a single special sort of knowledge on which the rest is based, one must have evidence for every bit of knowledge that is not of that most special sort. If one assumes that the question whether evidence is required is determined by what is claimed, independent of the particular circumstances in which it’s claimed, then one looks for evidence for any claim that might require evidence under any circumstance, however peculiar.

Thus, Warnock wants to know what my evidence is that I’m hearing a motor even when I’m standing next to a running car with the hood up, because it would be legitimate to wonder about evidence when I heard what seemed to be a car noise outside when I’m in a windowless room next to a special effects studio.

The idea that there is a special sort of sentence with a special property of not requiring evidence, quite independent of circumstances, is the origin of the idea of sense data. But it is wildly implausible.

Frog retinas have special “bug detector” neurons hard wired in: they fire when there is a small black dot moving erratically in the visual field. What sense data form the evidence for the frog’s belief that there is a bug? The question makes no sense, for it supposes that the frog sees something more basic than the bug and arrives at the bug belief “inferentially.” Insofar as the neurological evidence is relevant to what the frog sees, we have reason to believe that no inference is required.

Well, don’t we do something like inference, at least pretty often? Austin notes that we don’t arrive at material object language by starting with sense data language, we arrive at sense data language by backing away from material object language:

The evidence suggests that we have a good car detector, and our theory leads us to believe that it must be based on some other kind of detector that doesn’t require sticking out our necks quite so far. We have no experience of such a thing, our language includes no way to describe such a thing independent of the “riskier” claim, common sense and evolutionary biology suggest we are more likely to be built to detect cars than car-like sense data, and so forth.

When we try to build an artificially intelligent vacuum that can maneuver around the house, we attempt to build it along the lines of the foundationalist philosophical theory, thereby seeming to give the theory some new point. The inputs might be unanalyzed data about light intensities at various points on a two-dimensional field.

Ten or fifteen years ago, computer synthesized speech seemed to an impossible feat: The same phoneme gets produced in completely different ways depending on what phonemes come before and after. Here’s how the problem was finally solved: modern speech synthesis chips simulate the throat, vocal cords, teeth, tongue, and lips. You always move you mouth the same way to produce the same phoneme.

It seems reasonable to suspect that something similar may be true about object detection: the retina is not a passive recorder of impinging light like our current cameras are. At the very least, it also does some contrast enhancement and edge detection.

Though nothing comparable is known about vision, our eardrums vibrate because of external sound plus internal forces generated by, in some sense, what our brains "expect." That makes good engineering sense: if "what we expect" is normally more or less correct, it cancels out the external sound, and so we can detect a given sound with much less motion of the eardrum. Therefore we have a range of hearing that is much larger for a given flexibility of the eardrum. How does anyone know this? A crazy audiologist stuck a microphone in the ear of someone who was complaining that his ears were ringing. The audiologist was shocked to discover that he could hear it too. My wife is a psychiatrist, and every couple of years I try to persuade her to stick a microphone in the ear of someone having auditory hallucinations. But she, quite appropriately, won't do it. But the point that something like that is even conceivable shows that the idea that, even at the engineering level or the neurophysiological level, the counterpart of the foundationalist picture, figuring out how to reconstitute our experienced world from the physical data that we take to impinge on our sensory surfaces, is not obviously the right way to proceed, or the right picture to have in mind.

When an adult blind from birth is given vision surgically, they only learn to interpret any visual data at all with great difficulty, and, and this is my point, the color data remains completely disconnected from the rest of the visual data, at the very least, for a long time. The standard explanation is that color is handled in a different region of the brain from the rest of visual detection.

Foundationalism (about anything, not just the material world), as an unanalyzed presumption, is a disaster. That is the moral of our reading so far this semester. It is a disaster that has plagued philosophy at least since Descartes, and one we have only begun to escape in the past 50 years or so. (I think we have escaped it about the material world. About theoretical objects of science, abstract objects of mathematics, color, morality, fundamental constitution of the world, aesthetics, political philosophy (social contract theory), and surely lots of other things that haven't occurred to me, not so much.) That makes it an exciting time to be a philosopher: you aren’t looking at problems that have been gone over for two millennia, but ones that have only been around for 50 years.

In addition, and I am here discussing a new point I've made today, analogs of the foundational view are assumed in all sorts of scientific and engineering contexts (atomism, object recognition from patterns of impinging light, compositional semantics) usually without clear recognition that there are other options ready to hand. Thus, there are whole families of possibilities that don't get the attention that they deserve.

-- ShaughanLavine - 23 April 2003 - 17 Apr 2007