artiﬁcial intelligence -...

Artificial IntelligenceFoundations of Computational Agents

Artificial Intelligence: Foundations of Computational Agents is about the science ofartificial intelligence (AI). It presents AI as the study of the design of intelligent com-putational agents. The book is structured as a textbook, but it is accessible to a wideaudience of professionals and researchers.

The past decades have witnessed the emergence of AI as a serious science andengineering discipline. This book provides the first accessible synthesis of the fieldaimed at undergraduate and graduate students. It provides a coherent vision of thefoundations of the field as it is today, in terms of a multidimensional design spacethat has been partially explored. As with any science worth its salt, AI has a coherent,formal theory and a rambunctious experimental wing. The book balances theory andexperiment, showing how to link them intimately together. It develops the science ofAI together with its engineering applications.

David L. Poole is Professor of Computer Science at the University of British Columbia.He is a coauthor of Computational Intelligence: A Logical Approach (1998), cochair of theTwenty-Fourth AAAI Conference on Artificial Intelligence (AAAI-10), and coeditorof the Proceedings of the Tenth Conference in Uncertainty in Artificial Intelligence (1994).Poole is a former associate editor of the Journal of Artificial Intelligence Research. He isan associate editor of Artificial Intelligence and on the editorial boards of AI Magazineand AAAI Press. He is the secretary of the Association for Uncertainty in ArtificialIntelligence and is a Fellow of the Association for the Advancement of ArtificialIntelligence.

Alan K. Mackworth is Professor of Computer Science and Canada Research Chairin Artificial Intelligence at the University of British Columbia. He has authored morethan 100 papers and coauthored the text Computational Intelligence: A Logical Approach.He was President and Trustee of International Joint Conferences on AI (IJCAI) Inc.Mackworth was vice president and president of the Canadian Society for Compu-tational Studies of Intelligence (CSCSI). He has served as president of the AAAI.He also served as the founding director of the UBC Laboratory for ComputationalIntelligence. He is a Fellow of the Canadian Institute for Advanced Research, AAAI,and the Royal Society of Canada.

www.cambridge.org© in this web service Cambridge University Press

Cambridge University Press978-0-521-51900-7 - Artificial Intelligence: Foundations of Computational AgentsDavid L. Poole and Alan K. MackworthFrontmatterMore information

http://www.cambridge.org/9780521519007

http://www.cambridge.org


Artificial Intelligence

Foundations of Computational Agents

David L. Poole

University of British Columbia

Alan K. Mackworth

University of British Columbia






CAMBRIDGE UNIVERSITY PRESSCambridge, New York, Melbourne, Madrid, Cape Town, Singapore,

Sao Paulo, Delhi, Dubai, Tokyo

Cambridge University Press32 Avenue of the Americas, New York, NY 10013-2473, USA

www.cambridge.orgInformation on this title: www.cambridge.org/9780521519007

C© David L. Poole and Alan K. Mackworth 2010

This publication is in copyright. Subject to statutory exceptionand to the provisions of relevant collective licensing agreements,no reproduction of any part may take place without the written

permission of Cambridge University Press.

First published 2010

Printed in the United States of America

A catalog record for this publication is available from the British Library.

Library of Congress Cataloging in Publication data

Poole, David L. (David Lynton), 1958–Artificial intelligence : foundations of computational agents / David L. Poole, Alan K. Mackworth.

p. cm.Includes bibliographical references and index.

ISBN 978-0-521-51900-7 (hardback)1. Computational intelligence – Textbooks. 2. Artificial intelligence – Textbooks. I. Mackworth, Alan K.

II. Title.Q342.P66 2010

006.3 – dc22 2009039895

ISBN 978-0-521-51900-7 Hardback

Cambridge University Press has no responsibility for the persistence oraccuracy of URLs for external or third-party Internet Web sites referred to in

this publication and does not guarantee that any content on such Web sites is,or will remain, accurate or appropriate.






To our families for their love, support, and patience

Jennifer, Alexandra, and Shannon

Marian and Bryn






Contents

Preface xiii

I Agents in the World: What Are Agents and How Can They BeBuilt? 1

1 Artificial Intelligence and Agents 31.1 What Is Artificial Intelligence? . . . . . . . . . . . . . . . . . . 31.2 A Brief History of AI . . . . . . . . . . . . . . . . . . . . . . . . 61.3 Agents Situated in Environments . . . . . . . . . . . . . . . . . 101.4 Knowledge Representation . . . . . . . . . . . . . . . . . . . . 111.5 Dimensions of Complexity . . . . . . . . . . . . . . . . . . . . . 191.6 Prototypical Applications . . . . . . . . . . . . . . . . . . . . . 291.7 Overview of the Book . . . . . . . . . . . . . . . . . . . . . . . 391.8 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401.9 References and Further Reading . . . . . . . . . . . . . . . . . 401.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2 Agent Architectures and Hierarchical Control 432.1 Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432.2 Agent Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442.3 Hierarchical Control . . . . . . . . . . . . . . . . . . . . . . . . 502.4 Embedded and Simulated Agents . . . . . . . . . . . . . . . . 592.5 Acting with Reasoning . . . . . . . . . . . . . . . . . . . . . . . 602.6 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

vii






viii Contents

2.7 References and Further Reading . . . . . . . . . . . . . . . . . 662.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

II Representing and Reasoning 69

3 States and Searching 713.1 Problem Solving as Search . . . . . . . . . . . . . . . . . . . . . 713.2 State Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723.3 Graph Searching . . . . . . . . . . . . . . . . . . . . . . . . . . 743.4 A Generic Searching Algorithm . . . . . . . . . . . . . . . . . . 773.5 Uninformed Search Strategies . . . . . . . . . . . . . . . . . . . 793.6 Heuristic Search . . . . . . . . . . . . . . . . . . . . . . . . . . . 873.7 More Sophisticated Search . . . . . . . . . . . . . . . . . . . . . 923.8 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1063.9 References and Further Reading . . . . . . . . . . . . . . . . . 1063.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

4 Features and Constraints 1114.1 Features and States . . . . . . . . . . . . . . . . . . . . . . . . . 1114.2 Possible Worlds, Variables, and Constraints . . . . . . . . . . . 1134.3 Generate-and-Test Algorithms . . . . . . . . . . . . . . . . . . 1184.4 Solving CSPs Using Search . . . . . . . . . . . . . . . . . . . . 1194.5 Consistency Algorithms . . . . . . . . . . . . . . . . . . . . . . 1204.6 Domain Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . 1254.7 Variable Elimination . . . . . . . . . . . . . . . . . . . . . . . . 1274.8 Local Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1304.9 Population-Based Methods . . . . . . . . . . . . . . . . . . . . 1414.10 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1444.11 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1514.12 References and Further Reading . . . . . . . . . . . . . . . . . 1514.13 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

5 Propositions and Inference 1575.1 Propositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1575.2 Propositional Definite Clauses . . . . . . . . . . . . . . . . . . 1635.3 Knowledge Representation Issues . . . . . . . . . . . . . . . . 1745.4 Proving by Contradictions . . . . . . . . . . . . . . . . . . . . . 1855.5 Complete Knowledge Assumption . . . . . . . . . . . . . . . . 1935.6 Abduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1995.7 Causal Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2045.8 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2065.9 References and Further Reading . . . . . . . . . . . . . . . . . 2075.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208






Contents ix

6 Reasoning Under Uncertainty 2196.1 Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2196.2 Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2326.3 Belief Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . 2356.4 Probabilistic Inference . . . . . . . . . . . . . . . . . . . . . . . 2486.5 Probability and Time . . . . . . . . . . . . . . . . . . . . . . . . 2666.6 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2746.7 References and Further Reading . . . . . . . . . . . . . . . . . 2746.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

III Learning and Planning 281

7 Learning: Overview and Supervised Learning 2837.1 Learning Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . 2847.2 Supervised Learning . . . . . . . . . . . . . . . . . . . . . . . . 2887.3 Basic Models for Supervised Learning . . . . . . . . . . . . . . 2987.4 Composite Models . . . . . . . . . . . . . . . . . . . . . . . . . 3137.5 Avoiding Overfitting . . . . . . . . . . . . . . . . . . . . . . . . 3207.6 Case-Based Reasoning . . . . . . . . . . . . . . . . . . . . . . . 3247.7 Learning as Refining the Hypothesis Space . . . . . . . . . . . 3277.8 Bayesian Learning . . . . . . . . . . . . . . . . . . . . . . . . . 3347.9 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3407.10 References and Further Reading . . . . . . . . . . . . . . . . . 3417.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342

8 Planning with Certainty 3498.1 Representing States, Actions, and Goals . . . . . . . . . . . . . 3508.2 Forward Planning . . . . . . . . . . . . . . . . . . . . . . . . . . 3568.3 Regression Planning . . . . . . . . . . . . . . . . . . . . . . . . 3578.4 Planning as a CSP . . . . . . . . . . . . . . . . . . . . . . . . . . 3608.5 Partial-Order Planning . . . . . . . . . . . . . . . . . . . . . . . 3638.6 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3668.7 References and Further Reading . . . . . . . . . . . . . . . . . 3678.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367

9 Planning Under Uncertainty 3719.1 Preferences and Utility . . . . . . . . . . . . . . . . . . . . . . . 3739.2 One-Off Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . 3819.3 Sequential Decisions . . . . . . . . . . . . . . . . . . . . . . . . 3869.4 The Value of Information and Control . . . . . . . . . . . . . . 3969.5 Decision Processes . . . . . . . . . . . . . . . . . . . . . . . . . 3999.6 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4129.7 References and Further Reading . . . . . . . . . . . . . . . . . 4139.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413






x Contents

10 Multiagent Systems 42310.1 Multiagent Framework . . . . . . . . . . . . . . . . . . . . . . . 42310.2 Representations of Games . . . . . . . . . . . . . . . . . . . . . 42510.3 Computing Strategies with Perfect Information . . . . . . . . . 43010.4 Partially Observable Multiagent Reasoning . . . . . . . . . . . 43310.5 Group Decision Making . . . . . . . . . . . . . . . . . . . . . . 44510.6 Mechanism Design . . . . . . . . . . . . . . . . . . . . . . . . . 44610.7 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44910.8 References and Further Reading . . . . . . . . . . . . . . . . . 44910.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450

11 Beyond Supervised Learning 45111.1 Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45111.2 Learning Belief Networks . . . . . . . . . . . . . . . . . . . . . 45811.3 Reinforcement Learning . . . . . . . . . . . . . . . . . . . . . . 46311.4 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48511.5 References and Further Reading . . . . . . . . . . . . . . . . . 48611.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 486

IV Reasoning About Individuals and Relations 489

12 Individuals and Relations 49112.1 Exploiting Structure Beyond Features . . . . . . . . . . . . . . 49212.2 Symbols and Semantics . . . . . . . . . . . . . . . . . . . . . . 49312.3 Datalog: A Relational Rule Language . . . . . . . . . . . . . . 49412.4 Proofs and Substitutions . . . . . . . . . . . . . . . . . . . . . . 50612.5 Function Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . 51212.6 Applications in Natural Language Processing . . . . . . . . . . 52012.7 Equality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53212.8 Complete Knowledge Assumption . . . . . . . . . . . . . . . . 53712.9 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54112.10 References and Further Reading . . . . . . . . . . . . . . . . . 54212.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542

13 Ontologies and Knowledge-Based Systems 54913.1 Knowledge Sharing . . . . . . . . . . . . . . . . . . . . . . . . . 54913.2 Flexible Representations . . . . . . . . . . . . . . . . . . . . . . 55013.3 Ontologies and Knowledge Sharing . . . . . . . . . . . . . . . 56313.4 Querying Users and Other Knowledge Sources . . . . . . . . . 57613.5 Implementing Knowledge-Based Systems . . . . . . . . . . . . 57913.6 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59113.7 References and Further Reading . . . . . . . . . . . . . . . . . 59113.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592






Contents xi

14 Relational Planning, Learning, and Probabilistic Reasoning 59714.1 Planning with Individuals and Relations . . . . . . . . . . . . 59814.2 Learning with Individuals and Relations . . . . . . . . . . . . 60614.3 Probabilistic Relational Models . . . . . . . . . . . . . . . . . . 61114.4 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61814.5 References and Further Reading . . . . . . . . . . . . . . . . . 61814.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620

V The Big Picture 623

15 Retrospect and Prospect 62515.1 Dimensions of Complexity Revisited . . . . . . . . . . . . . . . 62515.2 Social and Ethical Consequences . . . . . . . . . . . . . . . . . 62915.3 References and Further Reading . . . . . . . . . . . . . . . . . 632

A Mathematical Preliminaries and Notation 633A.1 Discrete Mathematics . . . . . . . . . . . . . . . . . . . . . . . . 633A.2 Functions, Factors, and Arrays . . . . . . . . . . . . . . . . . . 634A.3 Relations and the Relational Algebra . . . . . . . . . . . . . . . 635

Bibliography 637

Index 653






Preface

Artificial Intelligence: Foundations of Computational Agents is a book about thescience of artificial intelligence (AI). The view we take is that AI is the studyof the design of intelligent computational agents. The book is structured as atextbook, but it is designed to be accessible to a wide audience.

We wrote this book because we are excited about the emergence of AI as anintegrated science. As with any science worth its salt, AI has a coherent, formaltheory and a rambunctious experimental wing. Here we balance theory andexperiment and show how to link them intimately together. We develop thescience of AI together with its engineering applications. We believe the adage“There is nothing so practical as a good theory.” The spirit of our approachis captured by the dictum “Everything should be made as simple as possible,but not simpler.” We must build the science on solid foundations; we presentthe foundations, but only sketch, and give some examples of, the complexityrequired to build useful intelligent systems. Although the resulting systemswill be complex, the foundations and the building blocks should be simple.

The book works as an introductory text on AI for advanced undergrad-uate or graduate students in computer science or related disciplines such ascomputer engineering, philosophy, cognitive science, or psychology. It willappeal more to the technically minded; parts are technically challenging, fo-cusing on learning by doing: designing, building, and implementing systems.Any curious scientifically oriented reader will benefit from studying the book.Previous experience with computational systems is desirable, but prior studyof the foundations on which we build, including logic, probability, calcu-lus, and control theory, is not necessary, because we develop the concepts asrequired.

xiii






xiv Preface

The serious student will gain valuable skills at several levels ranging fromexpertise in the specification and design of intelligent agents to skills for imple-menting, testing, and improving real software systems for several challengingapplication domains. The thrill of participating in the emergence of a new sci-ence of intelligent agents is one of the attractions of this approach. The practicalskills of dealing with a world of ubiquitous, intelligent, embedded agents arenow in great demand in the marketplace.

The focus is on an intelligent agent acting in an environment. We start withsimple agents acting in simple, static environments and gradually increase thepower of the agents to cope with more challenging worlds. We explore ninedimensions of complexity that allow us to introduce, gradually and with mod-ularity, what makes building intelligent agents challenging. We have tried tostructure the book so that the reader can understand each of the dimensionsseparately, and we make this concrete by repeatedly illustrating the ideas withfour different agent tasks: a delivery robot, a diagnostic assistant, a tutoringsystem, and a trading agent.

The agent we want the student to envision is a hierarchically designedagent that acts intelligently in a stochastic environment that it can only par-tially observe – one that reasons about individuals and the relationships amongthem, has complex preferences, learns while acting, takes into account otheragents, and acts appropriately given its own computational limitations. Ofcourse, we can’t start with such an agent; it is still a research question to buildsuch agents. So we introduce the simplest agents and then show how to addeach of these complexities in a modular way.

We have made a number of design choices that distinguish this book fromcompeting books, including the earlier book by the same authors:

• We have tried to give a coherent framework in which to understand AI.We have chosen not to present disconnected topics that do not fit to-gether. For example, we do not present disconnected logical and prob-abilistic views of AI, but we have presented a multidimensional designspace in which the students can understand the big picture, in whichprobabilistic and logical reasoning coexist.

• We decided that it is better to clearly explain the foundations on whichmore sophisticated techniques can be built, rather than present thesemore sophisticated techniques. This means that a larger gap exists be-tween what is covered in this book and the frontier of science. It alsomeans that the student will have a better foundation to understand cur-rent and future research.

• One of the more difficult decisions we made was how to linearize thedesign space. Our previous book (Poole, Mackworth, and Goebel, 1998)presented a relational language early and built the foundations in termsof this language. This approach made it difficult for the students toappreciate work that was not relational, for example, in reinforcement






Preface xv

learning that is developed in terms of states. In this book, we have chosena relations-late approach. This approach probably reflects better the re-search over the past few decades in which there has been much progressin feature-based representations. It also allows the student to understandthat probabilistic and logical reasoning are complementary. The book,however, is structured so that an instructor can present relations earlier.

This book uses examples from AIspace.org (http://www.aispace.org), a col-lection of pedagogical applets that we have been involved in designing. Togain further experience in building intelligent systems, a student should alsoexperiment with a high-level symbol-manipulation language, such as LISP orProlog. We also provide implementations in AILog, a clean logic programminglanguage related to Prolog, designed to demonstrate many of the issues in thisbook. This connection is not essential to an understanding or use of the ideasin this book.

Our approach, through the development of the power of the agent’s capa-bilities and representation language, is both simpler and more powerful thanthe traditional approach of surveying and cataloging various applications ofAI. However, as a consequence, some applications, such as the details of com-putational vision or computational linguistics, are not covered in this book.

We have chosen not to present an encyclopedic view of AI. Not every ma-jor idea that has been investigated is presented here. We have chosen somebasic ideas on which other, more sophisticated, techniques are based andhave tried to explain the basic ideas in detail, sketching how these can beexpanded.

Figure 1 (page xvi) shows the topics covered in the book. The solid linesgive prerequisites. Often the prerequisite structure does not include all sub-topics. Given the medium of a book, we have had to linearize the topics. How-ever, the book is designed so that the topics can be taught in any order satisfy-ing the prerequisite structure.

The references given at the end of each chapter are not meant to be compre-hensive: we have referenced works that we have directly used and works thatwe think provide good overviews of the literature, by referencing both classicworks and more recent surveys. We hope that no researchers feel slighted bytheir omission, and we are happy to have feedback where someone feels thatan idea has been misattributed. Remember that this book is not a survey of AIresearch.

We invite you to join us in an intellectual adventure: building a science ofintelligent agents.

David PooleAlan Mackworth






xvi Preface

1: AI & Agents

2: Architecture & Control 3: States &

Searching

4: Features & Constraints

5: Propositions & Inference

6: Uncertainty

7: Supervised Learning

8: Planning

9: Planning Under Uncertainty

10: Multi agent systems

11: Beyond Supervised Learning

12: Individuals & Relations

13: Ontologies & KBS

14: Relational Planning Learning & Probability

Figure 1: Overview of chapters and dependencies






Preface xvii

Acknowledgments

Thanks to Randy Goebel for valuable input on this book. We also grate-fully acknowledge the helpful comments on earlier drafts of this book re-ceived from Giuseppe Carenini, Cristina Conati, Mark Crowley, PooyanFazli, Holger Hoos, Manfred Jaeger, Mohammad Reza Khojasteh, JacekKisynski, Bob Kowalski, Kevin Leyton-Brown, Marian Mackworth, GabrielMurray, Alessandro Provetti, Marco Valtorta, and the anonymous reviewers.Thanks to the students who pointed out many errors in earlier drafts.

Thanks to Jen Fernquist for the web site design, and to Tom Sgouros forhyperlatex fixes. We are grateful to James Falen for permission to quote hispoem on constraints. Thanks to our editor Lauren Cowles and the staff at Cam-bridge University Press for all their support, encouragement, and help. All themistakes remaining are ours.






artiﬁcial intelligence -...

Documents