Pieces of knowledge

One central question about processes in the mind is: How is knowledge represented internally in the mind ? At the end there are many different kinds of objects and relationships between them in our universe, and we can use language for most cases to control those internal representations. So, what would be a good model to unify all these things, or how do mental data structures look like and how does knowledge access with language work ?

When I was reading the PhD thesis of Push Singh about an architecture for reflective common sense thinking lately, I came to understand better what the "frame-idea" of Marvin Minsky is about. It might be well possible that the smallest pieces of knowledge our mental agencies use during reasoning are "frames", i.e. structures that consist of a basic typical form and that can be tuned with a set of pre-defined knobs to specific instants. For example, the word "car" should arouse at each listener this frame: Some metallic structure with a certain color and 4 wheels at the bottom. All additional variables can be tuned by asking questions to the story-telling person or letting our memories do this job. And recursively, wheels, windows and cockpit are frames as well, and will arouse also certain agents in our cerbral cortex for vision and sound. (Also interesting, our physical models of light could be seen as a collection of frames: photons (i.e. particles), waves, rays.)

There are two interesting directions to walk from this frame-concept, the first is how are frames connected to our language and speech agencies and second, what kind of frames are there. For the first point, I got to understand better Minsky's polyneme concept. Language can be seen as interface system to transfer some meaning, knowledge or frame into another mind. For this at least two sub-systems are necessary: First some system for recognizing some concept and second a system to bring the short-term memories into certain states (maybe similar to compare this with zipping and unzipping some packets of information for transport). The polyneme does this recognizing (k-lines do the memorizing). For example we all have acquired a collection of basic structures in our minds for descriping circles, boxes, colors, substance, etc. A polyneme arouses agents in these different agents in different realms. This parallel processing of meaning is the underlying idea of Minsky's statement something has only a meaning when it has multiple meanings: If we have only one way to represent a thing, our agencies will fail to "turn" the object in the mind, when we want to apply some different way of thinking to this.

The second part that I just found interesting about frames, and by reading Push Singh's work, is, that stories can be seen as collection of frames as well. And even a full story can be seen as just some basic frame that can be used by other stories or ways to think. When we deliberate about how to solve a problem, we might just re-activate an old story with its main actors, with its protagonists and antagonists, the different objects in the scene, and then look carefully in our mind how the problem was solved the last time. A huge amount of knowledge is transferred by story-telling.