Cognitive Science and Artificial Intelligence

The Representation Problem

The so-called representation problem is probably the key issue for the study of cognition. In a narrow understanding, the representation problem addresses the question how knowledge is encoded in a cognitive system (e.g. symbolic versus distributed representation). However, there are deeper and more philosophical aspects of the problem:

The Dynamical Systems Approach

In terms of systems theory, the cognitive scenario is usually seen as a system (i.e. the cognitive system) in interaction with its environment. In my cognitive scenario there are two "intersecting systems"; in addition to the cognitive system the interaction game between (parts of) the environment and (parts of) the cognitive system should be regarded as some kind of system as well. By this move it becomes possible to formulate the representation problem as a problem of part-whole relationship:

In which way can a part of a system (the cognitive system) contain phenomena brought forth by the whole system (the system that consists of the environment and the cognitive system)?

There is a general answer to this question hidden in dynamical systems theory. Taken's theorem of embedding says that the attractor of a (whole) dynamical system can be fully reconstructed on the basis of the time series of a single variable. This can be applied to my cognitive scenario as follows:

What is usually regarded as an entity of the outside world (e.g. a table), should be seen as a spatio-temporal pattern ("the table behavior") brought forth by the cognitive system and the outside world in interaction. The cognitive system as a part of this larger system contains this pattern in its knowledge about its own future states: I have a meaningful representation of the table when I know what my own next experience will be in response to an action (e.g. I know that I will see that the object, which I moved to the table and which I dropped afterwards, stays on the table).