In an earlier post in this series, I wrote:
In short, it is the problem of knowledge that needs to be explained, rather than the problem of experience.
So today I will begin discussing the question of knowledge.
My starting point is a kind of empiricism. That is to say, I take the view that we acquire knowledge through experience. Or, said differently, knowledge is not inherited. The empiricism of John Locke seems in about the right direction, though of course Locke left much unexplained. Locke talked about ideas, and I take that to be about the same as what we mean when we talk of concepts. The question of knowledge, for Locke appears to be one of how we acquire our ideas or concepts.
By the time we get to Hume, the discussion has changed. Empiricism, to Hume, seems to be a question of how we decide which statements are true. Now that’s a huge change. You cannot even have a statement until you have the necessary concepts. So an account of how we decide which statements are true will fall far short of explaining how we acquire concepts. My version of empiricism is closer to that of Locke than to that of Hume.
If we learn by experience, then perception is an important part of that experience. There seems to be a tradition of assuming that perception is passive. This view, that perception is passive, is often credited to Descartes, though it may be older. It’s a convenient view if you want to assume that acquiring knowledge is something similar to determining which sentences are true. You can just take perception to be a black box that delivers observations. And then you need some sort of account of how to accumulate observations and use that accumulation as the basis for knowledge. That leads to induction and perhaps to Bayesian epistemology. I find it implausible.
The AI version of passive perception, is that sensory cells are stimulated, giving raw data for the brain which is assumed to be a computer. This, too, seems unlikely to me. I would expect the mass of raw data from sensory stimulus to be something like the “blooming, buzzing confusion” mentioned by William James (see, for example, the SEP article).
If we take raw sensory data as input, we might think of attempting to develop an explanatory theory to explain that data. There is a simple explanatory theory. Namely, that the body is covered with sensory cells that periodically emit data. That seems to be a complete and true theory. It is hard to see how we could find a better theory than that.
That, in a nutshell, is the poverty of stimulus problem for perception. There is a complete theory which does not mention the external world. If there is also a theory which does mention the external world, then possible theories must be vastly underdetermined by the data. So that gives us little basis for determining what the world is like.
Science is as well known source of knowledge. Scientists do not just wait for passively received data. They actively seek data. They carry out experiments for the purpose of coming up with data. Science is sometime said to be hypothetico-deductive. That is, scientists are said to form hypotheses, make deductions from those hypotheses, then design experiments to test whether the deductive conclusions hold up. But, at best, that’s an oversimplification of how science works. There are many experiments done before there is an hypothesis, and sometimes before there is data.
The use of experimentation by science suggests a far more active role in probing the environment than the passive perception model would allow. And we see something similar with children who learn about their world by pulling and poking to see what happens.
Stimulus response systems
The passive perception idea fits the broad notion of stimulus-response. We receive inputs (stimulus) and we respond perhaps with immediate action or perhaps by changing beliefs (as the Bayesian model might suggest). But I want to turn stimulus-response on its head. We stimulate the environment, and then we see how the environment responds. That gives a better account of the use of experimentation by scientists, and the use of poking and pulling by children.
It also better fits the needs of a biological organism. If the organism is to survive, then it needs water, air, food, etc. The organism can attempt to stimulate the environment, and see whether the environment responds by meeting some of the needs of the organism.
The idea, then, is that an organism can experiment with stimulating the environment, and attempt to learn which forms of stimulation work best. That gives a basis for pragmatic judgment as a measure of how well the response of the environment meets the needs of the organism. And we might develop a conceptualization of the world around us as we attempt to systematically organize our ways of stimulating the environment.
With passive perception, we would have to determine what the world is like based only on the data we receive. But once we allow the kind of interaction I have suggested, then our information about the world consists of both the data received and the actions that we performed in order to stimulate the environment to provide that data. So we have far more information about the world. This goes a long way toward solving the “poverty of stimulus” problem.
I have explained why I favor an active, rather than passive perception. In future posts I shall be discussing how we go about getting useful information about the world. That is to say, I will be attempting to address some of the principles involved in being able to perceive.