My work has demonstrated that four tasks are sufficient to explain the understanding of the novel concept continuum; the specific claim being made is that THE INTEGRATION OF MEMORY RETRIEVAL, ANALOGICAL MAPPING, BASE-CONSTRUCTIVE ANALOGY, AND PROBLEM REFORMULATION ARE SUFFICIENT AND NECESSARY FOR THE UNDERSTANDING OF THE NOVEL CONCEPT CONTINUUM. The sufficiency of the tasks will be shown by demonstrating that the computational model which embodies them is capable of performing creative understanding. The necessity side of the argument is more difficult. Since necessity is impossible to prove through computational means, I will, instead, provide an argument outlying precisely why the functionality contained in the four tasks is necessary. So, while I will not directly show that these specific four tasks are required, I will demonstrate (through argumentation) that the functionality which is captured by these specific four tasks is necessary. Any theory of creative understanding would need to incorporate that functionality.
Returning to the integration of the four tasks, later chapters will show that the four are combined in an iterative fashion. If more ``cycles'' through the process are allowed, more creative understanding can result. Unfortunately, there is a danger in this approach that unbounded iterations will lead to bizarre understanding resulting, where bizarre is defined as an understanding of some novel concept which proves non-useful. For example, if you see a Star Trek phaser for the first time and hypothesize that it is a device for keeping drinks cold without ice, you have, in some sense, understood a novel concept. However, the understanding does not allow you to make accurate predictions or answer questions about the device since it is incorrect. I would, therefore, classify this as a bizarre understanding.
I propose that THE ONTOLOGICAL REQUIREMENTS AND THE READING DOMAIN PROVIDE SUFFICIENT CONSTRAINTS ON THE UNDERSTANDING PROCESS. In order to understand novel concepts, it is often necessary to alter existing concepts. The ontology serves to restrict how these alterations may occur, thereby limiting the potential bizarreness which can result. The reading domain itself provides another bound on the process of understanding. A reasoner can stop attempting to understand a novel concept as soon as an understanding results which allows the reading of the text to continue. The understanding may need to be revised later if it is incorrect; this is acceptable as we are assuming the author will eventually provide the reader with enough information to produce the correct understanding. By taking this conservative approach as to when to stop the understanding cycle, the possibility of bizarre outcomes is greatly limited.
In a similar fashion, the fact that I am dealing with understanding situated in the domain of reading provides valuable constraints on other hard problems. Most notable is that of when to perform problem reformulation. In everyday interaction with the world, it is often difficult to tell exactly when we should discard an understanding--is it incorrect or do we simply not have enough knowledge of the situation to correctly utilize it? In reading, since the communicative agreement is assumed to exist, I theorize that problem reformulation is explicitly signaled by understanding failures. In other words, THE READING PROCESS CONSTRAINS THE PROBLEM OF IDENTIFYING UNDERSTANDING FAILURES, THEREBY ALLOWING PROBLEM REFORMULATION TO WORK.