The Necessity of Ordinary Experience

The Necessity of Ordinary Experience Robin Flanagan (FlanaganR@wcsu.edu) Department of Psychology, WCSU, 181 White Street Danbury, CT 06810 USA interpret on its own (see (Brooks, 2008) for example); it only means that silicon-based machinery isn’t necessarily constrained by the same physical qualities that constrain biological organisms. A lot of very interesting work in artificial intelligence, does examine cognition while taking biological constraints into consideration, and these lines of research have been extremely fruitful, which should help to support the idea that implementation does indeed matter. The embodied cognition paradigm already assumes, however, that implementation is a critical element of any intelligent system. Abstract I argue in this paper that ordinary experience is not only a nice part of everyday life; it is a necessity for the development of human knowledge. I begin by looking at why the particular biological machinery that defines our nervous system matters. I then examine the particular machineries that constrain but also foster the development of human knowledge. Finally, I examine the kinds of activities that foster the development of knowledge, given the constraints of the given machinery, and conclude that activities that are repeated often and that involve meaningful interaction with an inherently meaningful environment form a plausible basis for the formation of knowledge within the particular neural net machinery that evolution has produced for us. The Basic Machinery Keywords: Learning; neural networks; embodied cognition; practice; education; development; instructional technology That leads to the examination of the actual elements of the biological machinery from which the nervous system is constructed. There are, of course, very few elements in the biological machinery. The main element is an ordinary neuron, which is not too dissimilar from other cells in the biological organism. Like other cells in the biological organism the neuron is best at responding to elements in the immediate surroundings. In other words, the neuron is best at noticing what’s in its immediate neighborhood and responding by secreting to its immediate neighborhood. However, the neuron can take on very unusual shapes, and these shapes, make them particularly good for communicating with each other, by redefining what is meant by “its immediate neighborhood”. The maximized surface area of the neuron (the dendrites) allows the neuron to receive multiple messages simultaneously from other neurons or from the environment. The other part of the neuron’s unusual shape (the axon) can sometimes be quite a long extension of the cell body. The axon is the main tool that the neuron has at its disposal for communicating to other neurons or to the muscles. So just by changing its shape the neuron has the ability to get information from, and have an effect on, parts of the nervous system and ultimately parts of the body that are not apparently in its immediate neighborhood. This is important because the main technique that neurons have for getting information, and for sending information, involves the idea of simple local processing. So it’s important to note that “local” for the neuron has been redefined to include connections to quite distant elements of the nervous system and the biological organism. In fact, in the case of the photoreceptors, “local” involves light waves arriving in the immediate vicinity from potentially extremely distant locations. Simple local processing is the kind of processing that single-celled organisms developed at the very beginning of organized life, to detect things in their Mind and world in short have been evolved together, and in consequence are something of a mutual fit. (James, 1948, p. 4) The Implementation Problem The implementation question is the notion that once a system of knowledge has been completely and accurately articulated, it shouldn’t matter in what kind of machinery the system is implemented. This was a major assumption of cognitive science for quite a long time, and to its credit it was a very useful and fruitful assumption. If we assume that there is no important difference between carbon-based machinery and silicon-based machinery, and this is a very reasonable assumption, we can investigate and test knowledge systems on silicon-based machinery, machinery which is much easier to control, much easier to completely specify, and much easier to manipulate in ethical ways. However, this assumption has two gaping holes in it: how does the knowledge get into the machinery (most biological organisms have no programmers to install useful data structures or programs, while most silicon-based machines do have programmers), and how does the knowledge get interpreted (most silicon-based machines have intelligent “users” to interpret the output; most biological organisms must interpret the knowledge for themselves). If, instead of ignoring implementation, we examine how the actual machinery works, we find that there are many important constraints derived directly from the machinery that actually help us to understand how the knowledge gets incorporated into the machinery and how the “knowledge” in the system gets interpreted. This, of course, does not mean that a silicon-based machine couldn’t learn and