Reasoning about concepts

   Retrieving a concept from memory is not enough; in order to utilize a particular concept for the comprehension process, it must be reasoned about. This is achieved through the reasoning supertask. The outcome of this supertask is to produce an understanding of the concept that allows it to be used by other supertasks. In non-operational terms, understanding is achieved when the concept ``makes sense'' to the reasoner. In loose computational terms, this occurs when the reasoner is able to see how the concept in question would ``fit'' into existing knowledge. This fit can be achieved in a number of ways. One possibility is that the reasoner understands what the concept may possibly be used for later in the reasoning; this is the task of   prediction. Another possibility is that the reasoner understands the   function of a particular concept, often by doing ``backward'' reasoning upon seeing the kinds of activities the concept participates in. This task is referred to as   abduction. Finally, both prediction and abduction are used to ensure that a given concept is internally coherent; in other words, is there a valid explanation for how it achieves its current function. This understanding task is referred to as   explanation.

The mechanics of the understanding process are the subject of the next chapter, where creative understanding is described. Briefly, understanding must occur on each of the types of novel concepts which was described earlier, as well as on concepts which are ``completely'' known to the reasoner. For example, suppose I show a reasoner a six-inch, red pencil and explain that it is used for writing. If I later show the same reasoner the identical object, then the reasoner should be able to retrieve the concept and thereby understand the function of the pencil. This is a process I refer to as pure recognitional understanding; all that is necessary to arrive at an understanding is to recognize a prior concept. However, this does not mean that the understanding is always perfectly accurate. If, for example, the reasoner cannot retrieve the known concept from memory then this form of reasoning cannot occur. In Squire, for example, most readers will simply recognize the object of the Enterprise. This will enable them to predict possible future actions; for instance, when the Enterprise is trapped, it is often the case that Captain Kirk will attempt to free her. This is, therefore, a reasonable prediction to make about the story in general.

Another possibility which exists in texts is that the reader is familiar with a particular class of concepts but not with the exact instance being used in the story. The majority of readers will have some knowledge of squires--both the knightly helper and the gentleman's title. But, a reader may lack specific knowledge of the squire of Gothos, if only because Gothos is an unknown concept. Thus, although the squire of Gothos is a novel concept, it is not a radically novel one. In fact, it is an example of an instantiation novelty. The difference between pure recognitional understanding and the need for understanding I-novel concepts is a matter of what background knowledge is present in the reader and what knowledge can be retrieved. For example, many Star Trek fans will be able to understand the title of the story simply by retrieving the known concept of the squire of Gothos.

Concepts which are more ``novel'' to a given reasoner also require understanding, although the process is more involved. Consider the example story, Zoo. The zoo which is being discussed is not the standard concept of zoo. In reality, the zoo is more of a safari with zoo-like qualities. A reasoner can understand this concept through the use of analogical mapping if the proper base concepts exist in memory. The concept would be evolutionarily novel with respect to background knowledge, then. If existing concepts can be used as a base, then the reasoner is forced to understand a revolutionarily novel concept. If a reader knows about people and about normal, industrial robots, then the sentient robotic narrator of the story     Men Are Different ([#!story:men!#]) would be impossible to understand through analogical mapping (Chapter 6 explains why). However, it is possible to merge several concepts together to arrive at an understanding of the sentient robot--a process which goes well beyond the task of simply recognizing a known concept.