Leora Morgenstern

Inheritance comes of age: applying nonmonotonic techniques to problems in industry

Nonmonotonic reasoning is virtually absent from industry and has been so since its inception; the result is that the field is becoming marginalized within AI. I argue that this is because researchers in the area focus exclusively on commonsense problems which are irrelevant to industry and because few efficient algorithms and/or tools have been developed. A sensible strategy is thus to focus on industry problems and to develop solutions within tractable subtheories of nonmonotonic logic. I examine one of the few examples of nonmonotonic reasoning in industry -- inheritance of business rules in the medical insurance domain -- and show how the paradigm of inheritance with exceptions can be extended to a broader and more powerful kind of nonmonotonic reasoning. Finally I discuss the underlying lessons that can be generalized to other industry problems.

Mid-Sized Axiomatizations of Commonsense Problems: A Case Study in Egg Cracking

We present an axiomatization of a problem in commonsense reasoning, characterizing the proper procedure for cracking an egg and transferring its contents to a bowl. The axiomatization is mid-sized, larger than toy problems such as the Yale Shooting Problem or the Suitcase Problem, but much smaller than the comprehensive axiomatizations associated with CYC and HPKB. This size of axiomatization permits the development of non-trivial, reusable core theories of commonsense reasoning, acts as a testbed for existing theories of commonsense reasoning, and encourages the discovery of new problems in commonsense reasoning.We present portions of core theories of containment, falling, and pouring, integrated into Shanahans circumscriptive event calculus, and show how these can serve as the basis of an axiomatization that partly characterizes egg cracking. We discuss several commonsense reasoning problems encountered during this research, such as the Initial Specification Problem (a relative of the frame problem that occurs in theories in which fluents can trigger actions), and the Unobtainable State Problem (the problem of determining whether or not a theorem stating that one cannot get from one state to another is meaningful).

Motivated action theory: a formal theory of causal reasoning

When we reason about change over time, {\it causation} provides an implicit preference: we prefer sequences of situations in which one situation leads causally to the next, rather than sequences in which one situation follows another at random and without causal connections. In this paper, we explore the problem of temporal reasoning---reasoning about change over time---and the crucial role that causation plays in our intuitions. We examine previous approaches to temporal reasoning, and their shortcomings, in light of this analysis. We propose a new system for {\it causal reasoning}, motivated action theory, which builds upon causation as a crucial preference criterion. Motivated action theory solves the traditional problems of both forward and backward reasoning, and additionally provides a basis for a new theory of explanation.

An architecture of diversity for commonsense reasoning

Although computers excel at certain bounded tasks that are difficult for humans, such as solving integrals, they have difficulty performing commonsense tasks that are easy for humans, such as understanding stories. In this Technical Forum contribution, we discuss commonsense reasoning and what makes it difficult for computers. We contend that commonsense reasoning is too hard a problem to solve using any single artificial intelligence technique. We propose a multilevel architecture consisting of diverse reasoning and representation techniques that collaborate and reflect in order to allow the best techniques to be used for the many situations that arise in commonsense reasoning. We present story understanding—specifically, understanding and answering questions about progressively harder children’s texts—as a task for evaluating and scaling up a commonsense reasoning system.

A First-order Theory of Communication and Multi-agent Plans

This paper presents a theory expressed in first-order logic for describing and supporting inference about action, knowledge, planning, and communication, in an egalitarian multi-agent setting. The underlying ontology of the theory uses a situationbased temporal model and a possible-worlds model of knowledge. It supports plans and communications of a very general kind, both informative communications and requests. Communications may refer to states of the world or states of knowledge in the past, present, or future. We demonstrate that the theory is powerful enough to represent several interesting multi-agent planning problems and to justify their solutions. We have proven that the theory of knowledge, communication, and planning is consistent with a broad range of physical theories, despite the existence of a number of potential paradoxes.

Introduction: progress in formal commonsense reasoning

This special issue consists largely of expanded and revised versions of selected papers of the Fifth International Symposium on Logical Formalizations of Commonsense Reasoning (Common Sense 2001), held at New York University in May 2001.1,2 The Common Sense Symposia, first organized in 1991 by John McCarthy and held roughly biannually since, are dedicated to exploring the development of formal commonsense theories using mathematical logic. Commonsense reasoning is a central part of human behavior; no real intelligence is possible without it. Thus, the development of systems that exhibit commonsense behavior is a central goal of Artificial Intelligence. It has proven to be more difficult to create systems that are capable of commonsense reasoning than systems that can solve “hard” reasoning problems. There are chess-playing programs that beat champions [5] and expert systems that assist in clinical diagnosis [32], but no programs that reason about how far one must bend over to put on one’s socks. Part of the difficulty is the all-encompassing aspect of commonsense reasoning: any problem one looks at touches on many different types of knowledge. Moreover, in contrast to expert knowledge which is usually explicit, most commonsense knowledge is implicit. One of the prerequisites to developing commonsense reasoning systems is making this knowledge explicit. John McCarthy [25] first noted this need and suggested using formal logic to encode commonsense knowledge and reasoning. In the ensuing decades, there has been much research on the representation of knowledge in formal logic and on inference algorithms to

Theory, Practice, and Applications of Rules on the Web

We present textual logic (TL), a novel approach that enables rapid semi-automatic acquisition of rich logical knowledge from text. The resulting axioms are expressed as defeasible higher-order logic formulas in Rulelog, a novel extended form of declarative logic programs. A key element of TL is textual terminology, a phrasal style of knowledge in which words/word-senses are used directly as logical constants. Another key element of TL is a method for rapid interactive disambiguation as part of logic-based text interpretation. Existential quantifiers are frequently required, and we describe Rulelog’s approach to making existential knowledge be defeasible. We describe results from a pilot experiment that represented the knowledge from several thousand English sentences in the domain of college-level cell biology, for purposes of question-answering.

Planning, Executing, and Evaluating the Winograd Schema Challenge

The Winograd Schema Challenge was proposed by Hector Levesque in 2011 as an alternative to the Turing Test. Chief among its features is a simple question format that can span many commonsense knowledge domains. Questions are chosen so that they do not require specialized knoweldge or training, and are easy for humans to answer. This article details our plans to run the WSC and evaluate results.

AI at IBM Research

IBM has played an active role in AI research since the fields inception more than 50 years ago. In a trend that reflects the increasing demand for applications that behave intelligently, IBM today carries out most AI research in an interdisciplinary fashion by combining AI technology with other computing techniques to solve difficult technical problems. This article reports on the range of AI activities within IBM Research and discusses emerging issues. AI at IBM computer science research takes place in four broad areas: knowledge representation and reasoning; statistical AI; vision; and game playing.

Book reviews: Knowledge representation and commonsense reasoning: Reviews of four books

Ronald J. Brachman, Hector J. Levesque, Knowledge Representation and Reasoning, Morgan Kaufmann,ISBN 1558609326, 2004, 381 pages.Raymond Reiter, Knowledge in Action: Logical Foundations for Specifying and Implementing DynamicalSystems, MIT Press, ISBN 0262182181, 2001, 448 pages.Erik T. Mueller, Commonsense Reasoning, Morgan Kaufmann, ISBN 0123693888, 2006, 432 pages.Chitta Baral, Knowledge Representation, Reasoning and Declarative Problem Solving, Cambridge UniversityPress, ISBN 0521818028, 2003, 544 pages.