Errata for Reasoning about Rational Agents
This page summarises the known typos/bugs, ommissions, and
clarifications for RaRA. If you come across any more, please let me know.
Second Printing
I managed to correct some typos for the second printing of the book
(March 2002): here are the corrected pages
for this printing.
Typos/Bugs
- Chapter 4, pages 76 and 79: I defined the semantics of
predicates via the function Phi with respect to time
points. They should have been defined with respect to
situations, i.e., the function Phi on p76 should be a function
Phi : Pred x W x T ->..., instead of Phi : Pred x T -> ...
and the semantic rule for predicates
on page 79 should be ``is a member of Phi(P,w,t)'', not
``is a member of Phi(P,t)''. (Corrected in 2nd printing.)
- Chapter 4, page 79: in the seventh semantic rule, ``Goal'' should
be ``Des''. This is a hangover from a previous version of
LaTeX macros; as far as I am aware, this is the only place
in the book that I used ``Goal'' instead of ``Des'', but
any other occurrences of ``Goal'' should also be read as ``Des''.
(Corrected in 2nd printing.)
- Chapter 4, page 85: there are two typos in the proof of
Theorem 4.3(3): line 5 from the bottom should read
``(M,V,w'',t) |= phi'', not ``(M,V,w'',t) |= (Bel i phi)'' and
similarly line 8 from the bottom. Thanks to Joxe Gaintzarain
Ibarmia for pointing this out.
- Chapter 4, page 88: in Theorem 4.8(1), there should be
an ``E'' (existential path quantifier) before
(Happens a')? on the right hand side of the implication.
(Because ``Happens'' is a path formula.) Thanks to Joxe Gaintzarain
Ibarmia for pointing this out.
(Corrected in 2nd printing.)
- Chapter 5, page 93: I ommitted a (very) important part of the
definition of subworlds. For w to be a subworld of
w', then w must contain fewer paths than w', but otherwise, the
two worlds are identical --- they agree on the interpretation
of predicates, constants, and BDI modalities.
(Corrected in 2nd printing.)
- Chapter 5, page 108. Last paragraph. It should be "the first
of these properties, (5.37)'' instead of ``the first of these
properties (5.25)'' Thanks to Joxe Gaintzarain Ibarmia for
pointing this out.
- Chapter 5, page 109: the text would read better if the 1st
sentence following formula (5.39) was deleted. Also, it should
be ``sub'' rather than ``sup'' in every row of Table
5.6. Thanks to Gordon Beavers, Henry Hexmoor, and Joxe
Gaintzarain Ibarmia for pointing this out.
(Corrected in 2nd printing.)
- Chapter 5, page 110: Table 5.7. BDI-R2: it should be ``Bel =>
Des => Int'' instead of ``Int => Des => Bel''. Thanks to Joxe
Gaintzarain Ibarmia for pointing this out.
(Corrected in 2nd printing.)
- Chapter 6, page 116. In the definition of reachability, the
1st world in the sequence of worlds should be w_0, not w_1. In
the last paragraph, third line, the proof should read ``w' in
B^w_t'' instead of ``w' in B^w'_t''. Thanks to Joxe
Gaintzarain Ibarmia for pointing this out.
- Chapter 7, page 137: In the third paragraph, the sentence
starting ``The answer I adopt here...'' is misleading. It
would be better to have written ``The answer I adopt here is
that the best you can do is to convince me that you intend
that I believe this state of affairs.''
- Chapter 8, page 151: In the sixth line of the proof of Theorem
8.2, the the ``exists alpha'' should be deleted from the
formula at the start of the line.
- Chapter 8, page 157: In the definition of FormTeam, an agent
doesn't need to inform the team that they can jointly achieve
the goal (they will already be aware of this). It suffices to
define FormTeam as {RequestTh i g alpha phi}, but a better
definition is (Bel i (JCan g phi))?; {RequestTh i g alpha phi}.
(the agent requests the grouo while believing that they can do
it).
- Chapter 8, page 159: In the definition of J-Attempt,
fourth line, it should read (Achvs alpha psi), not
(Achvs alpha;psi?). (Corrected in 2nd printing.)
Omissions
- Chapter 2: I omitted to mention the work of John Pollock,
on architectures for rational agents. Pollock's
OSCAR architecture was developed with some similar concerns in
mind to the BDI model. I should have cited John's book
Cognitive Carpentry (MIT Press, 1995).
- Chapter 8: I omitted to mention the work of Milind Tambe and
colleagues on the STEAM framework for cooperative problem
solving, which makes use of a joint-intention model for
cooperative problem solving. I should have cited
M. Tambe, 1997, ``Towards Flexible Teamwork'', In Journal
of Artificial Intelligence Research, Volume 7, Pages
83-124.
Clarifications
- Chapter 2, page 41: the graph doesn't quite show that what
happens when dynamism tends to a maximum. Thanks to Gordon
Beavers and Henry Hexmoor for raising this issue.
- Chapter 4: Time points.
Finally, a note on terminology. The term ``state formula'' is
as used in the (extensive) literature on branching temporal
logic. Since this literature is very well established, I chose
to leave it unchanged in the book, although I don't use states
as such. The terms ``situation'' and ``time point'' are from
the BDI literature (Rao and Georgeff), and for similar reasons
I chose not to change them in the book. As ever, terminology
is not ideal. Time points are emphatically not
intended to correspond to ``times of the day''. In fact, there
are comparatively few temporal logics that I am aware of which
attempt to do this. In fact, time points are somewhat more
like states. Thanks to Gordon Beavers and Henry Hexmoor for
raising this issue.
That's it for now.