This page contains the errors that we know about from the first edition of Poole, Mackworth and Goebel, Computational Intelligence: A Logical Approach, Oxford University Press, 1998. Copyright © Poole, Mackworth, Goebel, and Oxford University Press, 1998. All Rights reserved.

If you find an error or something you don't understand let us know.

To see what printing a book is, look at the first left-hand-side page with printing on it. (This is the page that contains copyright information and Library of Congress Cataloging-in-Publication Data). At the bottom of this page you will find numbers: 10 9 8 7 6... The last of these numbers specifies which printing you have.

Errata in third and subsequent printings

• Page 111 Exercise 3.18, "cs486" should be "cs489" and "cs340" should be "cs350".
• Page 153, Figure 4.8. Insert "D_X := ND_X"  in the "then" part of the "if" statement.
  

Errata in first and second printings

Chapter 1: Computational Intelligence and Knowledge

• Page 6, paragraph 2, "of of" should be "of".

Chapter 2: A Representation and Reasoning System

• Page 63 Exercise 2.1 (a) should read "Give a model of the knowledge base."

Chapter 3: Using Definite Knowledge

• Page 77 the query cs_or_math_course(C) should be preceded by "?".
• Page 80 line 9 "is absence" -> "its absence"

Chapter 4: Searching

• Page 155, end of second last paragraph: "and does it the number of variables" should be "and does it in time linear in the number of variables."

Chapter 5: Representing Knowledge

• Figure 5.3 (page 176). There should be an arc between datalog and definite clauses (not just an arrowhead).
• Exercise 5.2 (page 196). The c in Fred's representation should be x.

Chapter 6: Knowledge Engineering

• Page 220 The end-bar for the example should be at the bottom of the page.
• Page 221, first paragraph. "HOW" and "WHY" should be swapped. HOW moves from the root to the leaves, and WHY moves from the leaves to the root of the tree.
• Example 6.14 (page 223). "...they can ask how live(l1)...". The "live" should be "lit".
• Page 230, Sec 6.8. "than is provided by the meta-level, you can't rely on the all" should be "than is provided by the meta-level, you can't rely on all the"

Chapter 7: Beyond Definite Knowledge

• none

Chapter 8: Actions and Planning

• none

Chapter 9: Assumption-based Reasoning

• none

Chapter 10: Using Uncertain Knowledge

• Page 348 Sec. 10.2: "$current(w_1)\geq 1.3$ means there are more than $1.3$ amps flowing" should be "$current(w_1)\geq 1.3$ means there are at least $1.3$ amps flowing"
• Page 364 (Sec. 10.3) "Consider the wiring example of Figure \ref{powerl} (page \pageref{powerl}). Whether light $l_1$ is lit depends only on whether there is power in wire $w_0$ and whether light $l_1$ is ok. Other variables, such as the position of switch $s_1$, whether light $l_2$ is lit, or who the Queen of Canada is, are irrelevant. Whether light $l_1$ is lit depends only on whether there is power in wire $w_0$ and whether light $l_1$ is ok." Remove the last sentence.
• Example 10.13 In the sentence crossing pages 364 and 366, w₃ should be w₀. Thus to affect the light l₁, any random variable either needs to affect whether there is power in wire w₀ or affect whether light l₁ is ok.
• Page 384 "Lets initially assume..." should be "Let's initially assume..."
• Page 389 Sec 10.4 "Lets continue Example ..." should be "Let's continue Example ..."
• Example 10.25 (page 390). (Thanks to Bonnie Webber.) For the case where you check and report and see smoke, 10 should be subtracted from the values. Thus it should read:

  E(V | call and check and report and see_smoke) = -210
  E(V | ¬call and check and report and see_smoke)
     = -5010 * P(fire | report and see_smoke) - 10 * P(¬fire | report and see_smoke)
     = -5010 * 0.964 - 10 * 0.056
     = -4830
  

Chapter 11: Learning

• Figure 11.7 (page 413). The error for iteration 80 should be 0.179 (not 0.178).
• Example 11.16 (page 426). "...the probability of observing n A's in m independent trials..." should be "...the probability of observing a particular sequence of n A's in m independent trials...". It is, of course, a particular sequence that will be observed. This isn't the binomial distribution which is the probability of observing any sequence of n A's in m independent trials (for the binomial distribution the number in the book needs to be multiplied by the number of ways you can get n A's in m trials).
• Page 426 (first paragraph after the heading). "P(model|data)=1" should be "P(data|model)=1".
• Example 11.16 (page 427). (Thanks to Bonnie Webber.) This looks as though it has been run using a different database.

  Let decision tree d₀ be the tree with no internal nodes and predicts read with probability ½ (this is the most likely tree with no internal nodes given the data). Then:

  P(data | d0) = ½9 * ½9 = 0.00000149
  

  Let decision tree d_1a be the tree that splits on length, such that when length=long it predicts reads with probability 0 and when length=short it predicts reads with probability 9/11.

  P(data | d1a) = 17 * (9/11)9 * (2/11)2 = 0.00543
  

  Let decision tree d_1b be the tree that splits on thread, such that when thread=new it predicts reads with probability 7/10 (as 7 out of the 10 examples with thread=new have user_action=reads) and when thread=follow_up it predicts reads with probability 2/8.

  P(data | d1b) = (7/10)7 * (3/10)3 * (6/8)6 * (2/8)2 =  0.0000247
  

Chapter 12: Building Situated Robots

• none

Appendix A: Glossary

• Consistent (p. 465) "can" should be "cannot". A set of beliefs is consistent if false cannot be derived.

Appendix B: The Prolog programming language

• none

Appendix C: Some more implemented systems

• Page 509 line 6, "had" should be "has"