The automation of reasoning with incomplete information. From semantic foundations to efficient computation (Q1387257)

In spite of the seemingly more general title, this book is largely about one specific approach to reasoning with incomplete information: default logic, especially constrained default logic proposed by the author. After a brief introduction to possible approaches (in Ch. 1) and to classical default logic (in Ch. 2), the author motivates and introduces his approach (in Ch. 3). The main reason for this approach is illustrated by the following example: suppose that we have several robots, and the \(i\)-th robot is usable (\(u_i\)) if it is not broken (\(\neg b_i\)). In default theory, this is described by a default \(:\neg b_i/u_i\), meaning, crudely speaking, that if we cannot conclude that the \(i\)-th robot is broken, then it is usable. If we know that one of the two robots is broken (\(b_1\vee b_2\)), but we do not know which of them is broken, then, in classical default theory, we cannot conclude that the \(i\)-th robot is broken and therefore, by using the above default, we are able to conclude (contrary to common sense) that both robots are usable. Constrained default logic, crudely speaking, invokes the defaults' conclusions only if these conclusions are globally consistent. In Ch. 4, the author shows how different versions of constrained default logic relate to other known modifications of default logic. Further additions to default logic include explicit notion of a context (Ch. 5), explicit information about which rules are more specific (Ch. 7), and which ``lemmas'' we want to add (Ch. 8). The author's definitions are justified (in Ch. 8) by a natural possible-world interpretation of incomplete knowledge. The most practically interesting part of the book is Ch. 9, in which the author presents an innovative algorithm for query-answering in default logics. A (somewhat more traditional) Prolog-based approach is outlined in Ch. 10.