![Page 1: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/1.jpg)
Towards a Theory of IntentionRevision
Workshop on the Dynamics of Intention andPreference
Eric Pacuit
Stanford University
February 27, 2008
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 1
![Page 2: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/2.jpg)
Plan
I Existing literature
I Underlying ETL model
I Elements of a theory of intention revision
I Many agents
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 2
![Page 3: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/3.jpg)
Some Literature
Stemming from Bratman’s planning theory of intention a numberof logics of rational agency have been developed:
I Cohen and Levesque; Rao and Georgeff (BDI); Meyer, van derHoek (KARO); Bratman, Israel and Pollack (IRMA); andmany others.
Some common features
I Underlying temporal model
I Belief, Desire, Intention, Plans, Actions are defined withcorresponding operators in a language
J.-J. Meyer and F. Veltman. Intelligent Agents and Common Sense Reasoning.Handbook of Modal Logic, 2007.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 3
![Page 4: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/4.jpg)
Some Literature
Stemming from Bratman’s planning theory of intention a numberof logics of rational agency have been developed:
I Cohen and Levesque; Rao and Georgeff (BDI); Meyer, van derHoek (KARO); Bratman, Israel and Pollack (IRMA); andmany others.
Some common features
I Underlying temporal model
I Belief, Desire, Intention, Plans, Actions are defined withcorresponding operators in a language
J.-J. Meyer and F. Veltman. Intelligent Agents and Common Sense Reasoning.Handbook of Modal Logic, 2007.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 3
![Page 5: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/5.jpg)
Intention Revision
Many of the frameworks do discuss some form of intention revision.
W. van der Hoek, W. Jamroga and M. Wooldridge. Towards a Theory ofIntention Revision. Synthese, 2007.
I Beliefs are sets of Linear Temporal Logic formulas (eg., ©ϕ)
I Desires are (possibly inconsistent) sets of Linear TemporalLogic formulas
I Practical reasoning rules: α← α1, α2, . . . , αn
I Intentions are derived from the agents current active plans(trees of practical reasoning rules)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 4
![Page 6: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/6.jpg)
Intention Revision
Many of the frameworks do discuss some form of intention revision.
W. van der Hoek, W. Jamroga and M. Wooldridge. Towards a Theory ofIntention Revision. Synthese, 2007.
I Beliefs are sets of Linear Temporal Logic formulas (eg., ©ϕ)
I Desires are (possibly inconsistent) sets of Linear TemporalLogic formulas
I Practical reasoning rules: α← α1, α2, . . . , αn
I Intentions are derived from the agents current active plans(trees of practical reasoning rules)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 4
![Page 7: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/7.jpg)
Intention Revision
Many of the frameworks do discuss some form of intention revision.
W. van der Hoek, W. Jamroga and M. Wooldridge. Towards a Theory ofIntention Revision. Synthese, 2007.
I Beliefs are sets of Linear Temporal Logic formulas (eg., ©ϕ)
I Desires are (possibly inconsistent) sets of Linear TemporalLogic formulas
I Practical reasoning rules: α← α1, α2, . . . , αn
I Intentions are derived from the agents current active plans(trees of practical reasoning rules)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 4
![Page 8: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/8.jpg)
Intention Revision
Many of the frameworks do discuss some form of intention revision.
W. van der Hoek, W. Jamroga and M. Wooldridge. Towards a Theory ofIntention Revision. Synthese, 2007.
I Beliefs are sets of Linear Temporal Logic formulas (eg., ©ϕ)
I Desires are (possibly inconsistent) sets of Linear TemporalLogic formulas
I Practical reasoning rules: α← α1, α2, . . . , αn
I Intentions are derived from the agents current active plans(trees of practical reasoning rules)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 4
![Page 9: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/9.jpg)
Intention Revision
Many of the frameworks do discuss some form of intention revision.
W. van der Hoek, W. Jamroga and M. Wooldridge. Towards a Theory ofIntention Revision. Synthese, 2007.
I Beliefs are sets of Linear Temporal Logic formulas (eg., ©ϕ)
I Desires are (possibly inconsistent) sets of Linear TemporalLogic formulas
I Practical reasoning rules: α← α1, α2, . . . , αn
I Intentions are derived from the agents current active plans(trees of practical reasoning rules)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 4
![Page 10: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/10.jpg)
Intention Revision
Many of the frameworks do discuss some form of intention revision.
W. van der Hoek, W. Jamroga and M. Wooldridge. Towards a Theory ofIntention Revision. Synthese, 2007.
I Beliefs are sets of Linear Temporal Logic formulas (eg., ©ϕ)
I Desires are (possibly inconsistent) sets of Linear TemporalLogic formulas
I Practical reasoning rules: α← α1, α2, . . . , αn
I Intentions are derived from the agents current active plans(trees of practical reasoning rules)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 4
![Page 11: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/11.jpg)
Intention Revision
Many of the frameworks do discuss some form of intention revision.
W. van der Hoek, W. Jamroga and M. Wooldridge. Towards a Theory ofIntention Revision. Synthese, 2007.
I Two types of beliefs: strong beliefs vs. weak beliefs (beliefsthat take into account the agent’s intentions)
I A dynamic update operator is defined ([Ω]ϕ)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 5
![Page 12: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/12.jpg)
Underlying ETL Model (single agent)
I N is a set of nodes, or states
I A is a set of primitive actions.
I is the successor relation on N (with the usual properties)
I l is a labeling function. Formally, l is a partial function fromN × N to A where l(n, n′) ∈ A if n n′ and undefinedotherwise.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 6
![Page 13: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/13.jpg)
Underlying ETL Model (single agent)
The agent’s uncertainty is represented by a relation ∼⊆ N × N:
I.e., n ∼ n′ if according to the agent’s current information (i.e., theevents the agent has observed), the agent cannot distinguish staten from n′
Some assumptions:
I Perfect Recall: If n ∼ n′, then hn = h′n. This means that theagents remembers all of its choices.
I No Miracles: If n ∼ n′ and there are n1 and n2 withl(n, n1) = l(n′, n2), then n1 ∼ n2.
I Uniform Actions: If n1 ∼ n2 and l(n1, n′) = a then there is a
n′′ such that l(n2, n′′) = a. This means the agents knows
which options are available.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 7
![Page 14: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/14.jpg)
Underlying ETL Model (single agent)
The agent’s uncertainty is represented by a relation ∼⊆ N × N:
I.e., n ∼ n′ if according to the agent’s current information (i.e., theevents the agent has observed), the agent cannot distinguish staten from n′
Some assumptions:
I Perfect Recall: If n ∼ n′, then hn = h′n. This means that theagents remembers all of its choices.
I No Miracles: If n ∼ n′ and there are n1 and n2 withl(n, n1) = l(n′, n2), then n1 ∼ n2.
I Uniform Actions: If n1 ∼ n2 and l(n1, n′) = a then there is a
n′′ such that l(n2, n′′) = a. This means the agents knows
which options are available.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 7
![Page 15: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/15.jpg)
Underlying ETL Model (single agent)
The agent’s uncertainty is represented by a relation ∼⊆ N × N:
I.e., n ∼ n′ if according to the agent’s current information (i.e., theevents the agent has observed), the agent cannot distinguish staten from n′
Some assumptions:
I Perfect Recall: If n ∼ n′, then hn = h′n. This means that theagents remembers all of its choices.
I No Miracles: If n ∼ n′ and there are n1 and n2 withl(n, n1) = l(n′, n2), then n1 ∼ n2.
I Uniform Actions: If n1 ∼ n2 and l(n1, n′) = a then there is a
n′′ such that l(n2, n′′) = a. This means the agents knows
which options are available.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 7
![Page 16: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/16.jpg)
Underlying ETL Model (single agent)
The agent’s uncertainty is represented by a relation ∼⊆ N × N:
I.e., n ∼ n′ if according to the agent’s current information (i.e., theevents the agent has observed), the agent cannot distinguish staten from n′
Some assumptions:
I Perfect Recall: If n ∼ n′, then hn = h′n. This means that theagents remembers all of its choices.
I No Miracles: If n ∼ n′ and there are n1 and n2 withl(n, n1) = l(n′, n2), then n1 ∼ n2.
I Uniform Actions: If n1 ∼ n2 and l(n1, n′) = a then there is a
n′′ such that l(n2, n′′) = a. This means the agents knows
which options are available.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 7
![Page 17: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/17.jpg)
Perfect Recall
t = 0
t = 1
t = 2
t = 3
e2 e4
e1 e5
e1 e3
e2 e3
e7 e6
e2 e1 e2
e4 e2
e1 e3
e7
i
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 8
![Page 18: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/18.jpg)
Perfect Recall
t = 0
t = 1
t = 2
t = 3
e2 e4
e1 e5
e1 e3
e2 e3
e7 e6
e2 e1 e2
e4 e2
e1 e3
e7
i
i
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 8
![Page 19: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/19.jpg)
Perfect Recall
t = 0
t = 1
t = 2
t = 3
e2 e4
e1 e5
e1 e3
e2 e3
e7 e6
e2 e1 e2
e4 e2
e1 e3
e7
i
i
i
i
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 8
![Page 20: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/20.jpg)
No Miracles
t = 0
t = 1
t = 2
t = 3
e2 e4
e1 e5
e1 e3
e2 e3
e7 e6
e2 e1 e2
e4 e2
e1 e3
e7i
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 9
![Page 21: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/21.jpg)
No Miracles
t = 0
t = 1
t = 2
t = 3
e2 e4
e1 e5
e1 e3
e2 e3
e7 e6
e2 e1 e2
e4 e2
e1 e3
e7i
i
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 9
![Page 22: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/22.jpg)
Uniform Actions
Uniform Actions: If n1 ∼ n2 and l(n1, n′) = a then there is a n′′
such that l(n2, n′′) = a. This means the agents knows which
options are available.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 10
![Page 23: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/23.jpg)
Histories
A history is a sequence of events or actions (for each node there isa history from a root to that node).
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 11
![Page 24: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/24.jpg)
Two types of uncertainty
Given two finite histories h and h′,
h ∼i h′ means given the events i has observed, h and h′
are indistinguishable
e2 e4
e1 e5 e2 e3e1
i
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 12
![Page 25: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/25.jpg)
Two types of uncertainty
Given two maximal histories H and H ′,
agent i may be uncertain which of the two will be thefinal outcome.
e2 e4
e1 e5 e2 e3e1
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 12
![Page 26: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/26.jpg)
Elements of a theory of intention revision
Suppose A = a1, a2, . . . is a set of primitive actions andP = p1, p2, . . . a set of propositional variables.
L0 be the propositional language generated by P.
Our basic expressions:
I B(ϕ : a1, . . . an) is intended to mean “the agent believes ϕbecause he intends to do a1, and he intends to do a2, . . ., andhe intends to do an.”
I I (a : ϕ1, . . . , ϕm) is intended to mean “the agent intends todo a because he does not believe ¬ϕ1 and he does not believe¬ϕ2,. . . , and he does not believe ¬ϕm.”
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 13
![Page 27: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/27.jpg)
Elements of a theory of intention revision
Suppose A = a1, a2, . . . is a set of primitive actions andP = p1, p2, . . . a set of propositional variables.
L0 be the propositional language generated by P.
Our basic expressions:
I B(ϕ : a1, . . . an) is intended to mean “the agent believes ϕbecause he intends to do a1, and he intends to do a2, . . ., andhe intends to do an.”
I I (a : ϕ1, . . . , ϕm) is intended to mean “the agent intends todo a because he does not believe ¬ϕ1 and he does not believe¬ϕ2,. . . , and he does not believe ¬ϕm.”
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 13
![Page 28: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/28.jpg)
Elements of a theory of intention revision
Suppose A = a1, a2, . . . is a set of primitive actions andP = p1, p2, . . . a set of propositional variables.
L0 be the propositional language generated by P.
Our basic expressions:
I B(ϕ : a1, . . . an) is intended to mean “the agent believes ϕbecause he intends to do a1, and he intends to do a2, . . ., andhe intends to do an.”
I I (a : ϕ1, . . . , ϕm) is intended to mean “the agent intends todo a because he does not believe ¬ϕ1 and he does not believe¬ϕ2,. . . , and he does not believe ¬ϕm.”
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 13
![Page 29: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/29.jpg)
Elements of a theory of intention revision
Suppose A = a1, a2, . . . is a set of primitive actions andP = p1, p2, . . . a set of propositional variables.
L0 be the propositional language generated by P.
Our basic expressions:
I B(ϕ : a1, . . . an) is intended to mean “the agent believes ϕbecause he intends to do a1, and he intends to do a2, . . ., andhe intends to do an.”
I I (a : ϕ1, . . . , ϕm) is intended to mean “the agent intends todo a because he does not believe ¬ϕ1 and he does not believe¬ϕ2,. . . , and he does not believe ¬ϕm.”
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 13
![Page 30: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/30.jpg)
Elements of a theory of intention revision
Remark: We do not allowing the following formulas:
1. B(I (a1 : p) : a2): “The agent believes that [he intends to doa1 because he does not believe ¬p] because he intends to doa2.
2. I (a : Bp): “The agent intends to do a because he does notbelieve that ¬Bp.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 14
![Page 31: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/31.jpg)
Digression: Knowing/Believing for a reason
I S. Artemov has developed justification logics with a similarflavor: t : ϕ is intended to mean “the agent knows ϕ forreason t”.
I t is called a proof polynomial and is intended to represent aderivation of ϕ.
• There is an algebraic structure on the set of proof polynomials:t + s, !t, t · s
I Mel Fitting has developed a Kripke style semantics.M. Fitting. Logic of Proofs, Semantically. APAL, 2006.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 15
![Page 32: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/32.jpg)
Digression: Knowing/Believing for a reason
I S. Artemov has developed justification logics with a similarflavor: t : ϕ is intended to mean “the agent knows ϕ forreason t”.
I t is called a proof polynomial and is intended to represent aderivation of ϕ.
• There is an algebraic structure on the set of proof polynomials:t + s, !t, t · s
I Mel Fitting has developed a Kripke style semantics.M. Fitting. Logic of Proofs, Semantically. APAL, 2006.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 15
![Page 33: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/33.jpg)
Digression: Knowing/Believing for a reason
I S. Artemov has developed justification logics with a similarflavor: t : ϕ is intended to mean “the agent knows ϕ forreason t”.
I t is called a proof polynomial and is intended to represent aderivation of ϕ.
• There is an algebraic structure on the set of proof polynomials:t + s, !t, t · s
I Mel Fitting has developed a Kripke style semantics.M. Fitting. Logic of Proofs, Semantically. APAL, 2006.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 15
![Page 34: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/34.jpg)
Elements of a theory of intention revision
Any Action: Let ? be a new action symbol (not in A). Theintended interpretation of ? is “any action”.
I B(ϕ) =def B(ϕ :?)
I I (a) =def I (a : >).
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 16
![Page 35: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/35.jpg)
Elements of a theory of intention revision
Any Action: Let ? be a new action symbol (not in A). Theintended interpretation of ? is “any action”.
I B(ϕ) =def B(ϕ :?)
I I (a) =def I (a : >).
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 16
![Page 36: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/36.jpg)
Elements of a theory of intention revision
A belief set B is any set of expressions of the form ϕ : a1, . . . , an
where ϕ ∈ L0 and a1, . . . , an ∈ A satisfying the followingproperties:
1. If ϕ1 : a1, . . . , an ∈ B and ϕ2 : b1, . . . , bm ∈ B thenϕ1 ∧ ϕ2 : a1, . . . , an, b1, . . . , bm ∈ B.
2. ` ϕ1 → ϕ2 (in propositional logic) and ϕ1 : a1, . . . , an ∈ Bthen ϕ2 : a1, . . . , an ∈ B
3. ⊥ : a1, . . . , an 6∈ B
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 17
![Page 37: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/37.jpg)
Elements of a theory of intention revision
An intention set I is any set of expressions of the forma : ϕ1, . . . , ϕn where a ∈ A and ϕ1, . . . , ϕn ∈ L0 satisfying thefollowing properties:
1. If a : ϕ1, . . . , ϕn ∈ I and a : ψ1, . . . ψm ∈ I thena : ϕ1, . . . , ϕn, ψ1, . . . , ψm ∈ I
2. If a : ϕ1, . . . , ϕn ∈ I then∧
i ϕi is logically consistent.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 18
![Page 38: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/38.jpg)
Elements of a theory of intention revision
Intention-Belief Pairs: A pair (I,B) where I is an intention setand B is a belief set is called an intention-belief pair.
The main idea is that a pair (I,B) is intended to represent theagents current intentions and beliefs.
However, not every pair (I,B) will represent “coherent” intentionsand beliefs.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 19
![Page 39: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/39.jpg)
Elements of a theory of intention revision
A pair (I,B) is coherent provided:
1. The intentions are grounded in current beliefs: (if a : ϕ ∈ Ithere ¬ϕ :? 6∈ B)
2. There are no cycles (eg., there is no ϕ : a ∈ B with a : ϕ ∈ I)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 20
![Page 40: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/40.jpg)
Elements of a theory of intention revision
A pair (I,B) is coherent provided:
1. The intentions are grounded in current beliefs: (if a : ϕ ∈ Ithere ¬ϕ :? 6∈ B)
2. There are no cycles (eg., there is no ϕ : a ∈ B with a : ϕ ∈ I)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 20
![Page 41: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/41.jpg)
Elements of a theory of intention revision
A pair (I,B) is coherent provided:
1. The intentions are grounded in current beliefs: (if a : ϕ ∈ Ithere ¬ϕ :? 6∈ B)
2. There are no cycles (eg., there is no ϕ : a ∈ B with a : ϕ ∈ I)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 20
![Page 42: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/42.jpg)
Elements of a theory of intention revisionRevision operators:
1. Change/update the reason for believing ϕ: Suppose thatthe agent currently believes ϕ because he intends to do a.Updating by ϕ : b involves changing/adding a reason forbelieving ϕ.
2. Update/revise beliefs: The input is a formula ϕ and,following AGM, the agent changes its belief set Bappropriately. Note that in order to maintain coherency, wemay need to drop some intentions.
3. Weak-add an intention: Add an intention providedcoherency is maintained otherwise do not add the intention.
4. Strong-add an intention: Add the intention and changebelief/intention set appropriately.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 21
![Page 43: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/43.jpg)
Elements of a theory of intention revisionRevision operators:
1. Change/update the reason for believing ϕ: Suppose thatthe agent currently believes ϕ because he intends to do a.Updating by ϕ : b involves changing/adding a reason forbelieving ϕ.
2. Update/revise beliefs: The input is a formula ϕ and,following AGM, the agent changes its belief set Bappropriately. Note that in order to maintain coherency, wemay need to drop some intentions.
3. Weak-add an intention: Add an intention providedcoherency is maintained otherwise do not add the intention.
4. Strong-add an intention: Add the intention and changebelief/intention set appropriately.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 21
![Page 44: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/44.jpg)
Elements of a theory of intention revisionRevision operators:
1. Change/update the reason for believing ϕ: Suppose thatthe agent currently believes ϕ because he intends to do a.Updating by ϕ : b involves changing/adding a reason forbelieving ϕ.
2. Update/revise beliefs: The input is a formula ϕ and,following AGM, the agent changes its belief set Bappropriately. Note that in order to maintain coherency, wemay need to drop some intentions.
3. Weak-add an intention: Add an intention providedcoherency is maintained otherwise do not add the intention.
4. Strong-add an intention: Add the intention and changebelief/intention set appropriately.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 21
![Page 45: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/45.jpg)
Elements of a theory of intention revisionRevision operators:
1. Change/update the reason for believing ϕ: Suppose thatthe agent currently believes ϕ because he intends to do a.Updating by ϕ : b involves changing/adding a reason forbelieving ϕ.
2. Update/revise beliefs: The input is a formula ϕ and,following AGM, the agent changes its belief set Bappropriately. Note that in order to maintain coherency, wemay need to drop some intentions.
3. Weak-add an intention: Add an intention providedcoherency is maintained otherwise do not add the intention.
4. Strong-add an intention: Add the intention and changebelief/intention set appropriately.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 21
![Page 46: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/46.jpg)
Elements of a theory of intention revisionRevision operators:
1. Change/update the reason for believing ϕ: Suppose thatthe agent currently believes ϕ because he intends to do a.Updating by ϕ : b involves changing/adding a reason forbelieving ϕ.
2. Update/revise beliefs: The input is a formula ϕ and,following AGM, the agent changes its belief set Bappropriately. Note that in order to maintain coherency, wemay need to drop some intentions.
3. Weak-add an intention: Add an intention providedcoherency is maintained otherwise do not add the intention.
4. Strong-add an intention: Add the intention and changebelief/intention set appropriately.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 21
![Page 47: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/47.jpg)
Many Agents
E. Pacuit, R. Parikh and E. Cogan. The Logic of Knowledge Based Applica-tions. Knowledge, Rationality and Action (Synthese) 149: 311 - 341 (2006).
Issues: obligations, group obligations, knowledge, groupknowledge, default obligations, etc.
An agent’s obligations are often dependent on what the agentknows, and indeed one cannot reasonably be expected to respondto a problem if one is not aware of its existence.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 22
![Page 48: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/48.jpg)
Many Agents
E. Pacuit, R. Parikh and E. Cogan. The Logic of Knowledge Based Applica-tions. Knowledge, Rationality and Action (Synthese) 149: 311 - 341 (2006).
Issues: obligations, group obligations, knowledge, groupknowledge, default obligations, etc.
An agent’s obligations are often dependent on what the agentknows, and indeed one cannot reasonably be expected to respondto a problem if one is not aware of its existence.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 22
![Page 49: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/49.jpg)
Many Agents
E. Pacuit, R. Parikh and E. Cogan. The Logic of Knowledge Based Applica-tions. Knowledge, Rationality and Action (Synthese) 149: 311 - 341 (2006).
Issues: obligations, group obligations, knowledge, groupknowledge, default obligations, etc.
An agent’s obligations are often dependent on what the agentknows, and indeed one cannot reasonably be expected to respondto a problem if one is not aware of its existence.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 22
![Page 50: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/50.jpg)
Motivating Example
1. Uma is a physician whose neighbour is ill. Uma does notknow and has not been informed. Uma has no obligation (asyet) to treat the neighbour.
2. Uma is a physician whose neighbour Sam is ill. Theneighbour’s daughter Ann comes to Uma’s house and tellsher. Now Uma does have an obligation to treat Sam, orperhaps call in an ambulance or a specialist.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 23
![Page 51: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/51.jpg)
Motivating Example
1. Uma is a physician whose neighbour is ill. Uma does notknow and has not been informed. Uma has no obligation (asyet) to treat the neighbour.
2. Uma is a physician whose neighbour Sam is ill. Theneighbour’s daughter Ann comes to Uma’s house and tellsher. Now Uma does have an obligation to treat Sam, orperhaps call in an ambulance or a specialist.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 23
![Page 52: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/52.jpg)
Formalizing Example 1 & 2
t = 0
t = 1
t = 2
t = 3
v
m c
t c t c
c
m c
t c t c
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 24
![Page 53: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/53.jpg)
Formalizing Example 1 & 2
t = 0
t = 1
t = 2
t = 3
v
m c
t c t c
c
m c
t c t c
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 24
![Page 54: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/54.jpg)
Formalizing Example 1 & 2
t = 0
t = 1
t = 2
t = 3
v
m c
t c t c
λu(v) = λu(c)
u
c
m c
t c t c
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 24
![Page 55: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/55.jpg)
Formalizing Example 1 & 2
t = 0
t = 1
t = 2
t = 3
v
m c
t c t c
u
u
u
u
c
m c
t c t c
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 24
![Page 56: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/56.jpg)
Formalizing Example 1 & 2
t = 0
t = 1
t = 2
t = 3
v
m c
t c t c
S ¬S
H, 1 |= ¬KuS
u
c
m c
t c t c
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 24
![Page 57: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/57.jpg)
Formalizing Example 1 & 2
t = 0
t = 1
t = 2
t = 3
v
m c
t c t c
S ¬S
S ¬S
λu(vm) = λu(cm)
u
c
m c
t c t c
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 24
![Page 58: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/58.jpg)
Learning from the Protocol
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 25
![Page 59: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/59.jpg)
Learning from the Protocol
t = 0
t = 1
t = 2
t = 3
r t
wg t wg t
W W W ¬WR R ¬R ¬R
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 25
![Page 60: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/60.jpg)
Learning from the Protocol
t = 0
t = 1
t = 2
t = 3
r t
wg t wg t
W W W ¬WR R ¬R ¬R¬KW ∧ ¬KR
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 25
![Page 61: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/61.jpg)
Learning from the Protocol
t = 0
t = 1
t = 2
t = 3
!W !W !W
r t
wg t wg t
W W WR R ¬R
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 25
![Page 62: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/62.jpg)
Learning from the Protocol
t = 0
t = 1
t = 2
t = 3
!W !W !W
r t
wg t wg t
W W WR R ¬RKW ∧ ¬KR
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 25
![Page 63: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/63.jpg)
Learning from the Protocol
t = 0
t = 1
t = 2
t = 3
!W !W !W
r t
wg t wg t
W W WR R ¬R
r t
t t
W ¬WR ¬R
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 25
![Page 64: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/64.jpg)
Learning from the Protocol
t = 0
t = 1
t = 2
t = 3
!W !W !W
r t
wg t wg t
W W WR R ¬R
r t
t t
W ¬WR ¬R¬KW ∧ ¬KR¬KW ∧ ¬KR
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 25
![Page 65: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/65.jpg)
Learning from the Protocol
t = 0
t = 1
t = 2
t = 3
!W !W !W
r t
wg t wg t
W W WR R ¬R
!W
r t
t t
R
¬R
W
¬W
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 25
![Page 66: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/66.jpg)
Learning from the Protocol
t = 0
t = 1
t = 2
t = 3
!W !W !W
r t
wg t wg t
W W WR R ¬R
!W
r t
t t
W
¬R
R
¬W
KW ∧ KRKW ∧ ¬KR
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 25
![Page 67: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/67.jpg)
Actions
Assume a finite set, Act ⊆ Σ, of primitive actions.
Assume Act = ∪i∈ACi where Ci ∩ Cj 6= ∅ for i 6= j .
Given a finite global history H and a ∈ Act,
a(H) = H ′ | Ha H ′and H ′a global history
H, t |= [a]ϕ iff H ′, t + 1 |= ϕ for each H ′ ∈ a(H)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 26
![Page 68: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/68.jpg)
Actions
Assume a finite set, Act ⊆ Σ, of primitive actions.
Assume Act = ∪i∈ACi where Ci ∩ Cj 6= ∅ for i 6= j .
Given a finite global history H and a ∈ Act,
a(H) = H ′ | Ha H ′and H ′a global history
H, t |= [a]ϕ iff H ′, t + 1 |= ϕ for each H ′ ∈ a(H)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 26
![Page 69: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/69.jpg)
Actions
Assume a finite set, Act ⊆ Σ, of primitive actions.
Assume Act = ∪i∈ACi where Ci ∩ Cj 6= ∅ for i 6= j .
Given a finite global history H and a ∈ Act,
a(H) = H ′ | Ha H ′and H ′a global history
H, t |= [a]ϕ iff H ′, t + 1 |= ϕ for each H ′ ∈ a(H)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 26
![Page 70: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/70.jpg)
Knowledge Based Obligations
Actions
H, t |= [a]ϕ iff H ′, t + 1 |= ϕ for each H ′ ∈ a(H)
I When an action is performed, it is performed at the nextmoment of time.
I Only one agent can perform some action at any moment.
I If no agents perform an action, then nature performs a ‘clocktick’.
I Each agent knows when it can perform an action.(〈ai 〉> → Ki 〈ai 〉>)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 27
![Page 71: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/71.jpg)
Knowledge Based Obligations
Actions
H, t |= [a]ϕ iff H ′, t + 1 |= ϕ for each H ′ ∈ a(H)
I When an action is performed, it is performed at the nextmoment of time.
I Only one agent can perform some action at any moment.
I If no agents perform an action, then nature performs a ‘clocktick’.
I Each agent knows when it can perform an action.(〈ai 〉> → Ki 〈ai 〉>)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 27
![Page 72: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/72.jpg)
Knowledge Based Obligations
Actions
H, t |= [a]ϕ iff H ′, t + 1 |= ϕ for each H ′ ∈ a(H)
I When an action is performed, it is performed at the nextmoment of time.
I Only one agent can perform some action at any moment.
I If no agents perform an action, then nature performs a ‘clocktick’.
I Each agent knows when it can perform an action.(〈ai 〉> → Ki 〈ai 〉>)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 27
![Page 73: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/73.jpg)
Knowledge Based Obligations
Actions
H, t |= [a]ϕ iff H ′, t + 1 |= ϕ for each H ′ ∈ a(H)
I When an action is performed, it is performed at the nextmoment of time.
I Only one agent can perform some action at any moment.
I If no agents perform an action, then nature performs a ‘clocktick’.
I Each agent knows when it can perform an action.(〈ai 〉> → Ki 〈ai 〉>)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 27
![Page 74: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/74.jpg)
Knowledge Based Obligations
Actions
H, t |= [a]ϕ iff H ′, t + 1 |= ϕ for each H ′ ∈ a(H)
I When an action is performed, it is performed at the nextmoment of time.
I Only one agent can perform some action at any moment.
I If no agents perform an action, then nature performs a ‘clocktick’.
I Each agent knows when it can perform an action.(〈ai 〉> → Ki 〈ai 〉>)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 27
![Page 75: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/75.jpg)
Knowledge Based Obligations
Values: Informal Definition
All global histories will be presumed to have a value
Let G(H) be the set of extensions of (finite history) H which havethe highest possible value. (Assumptions are needed to make G(H)well defined)
We will say that a is good to be performed at H if G(H) ⊆ a(H),i.e., there are no H-good histories which do not involve theperforming of a.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 28
![Page 76: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/76.jpg)
Knowledge Based Obligations
Values: Informal Definition
All global histories will be presumed to have a value
Let G(H) be the set of extensions of (finite history) H which havethe highest possible value. (Assumptions are needed to make G(H)well defined)
We will say that a is good to be performed at H if G(H) ⊆ a(H),i.e., there are no H-good histories which do not involve theperforming of a.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 28
![Page 77: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/77.jpg)
Knowledge Based Obligations
Knowledge Based Obligation
Agent i has a (knowledge based) obliged to perform action a atglobal history H and time t iff a is an action which i (only) canperform, and i knows that it is good to perform a.
For each a ∈ Act, let G (a) be a formula:
H, t |= G (a) iff G(Ht) ⊆ a(Ht)
Then we say that i is obliged to perform action a (at H, t) ifKi (G (a)) is true (at H, t).
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 29
![Page 78: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/78.jpg)
Knowledge Based Obligations
Example 2
t = 0
t = 1
t = 2
t = 3
v
m c
t c t c
S ¬S
S ¬S
λu(vm) = λu(cm)
u
c
m c
t c t c
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 30
![Page 79: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/79.jpg)
Knowledge Based Obligations
Example 2
t = 0
t = 1
t = 2
t = 3
v
m c
t c t c
S ¬S
S ¬S
2 1 1 2
u
c
m c
t c t c
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 30
![Page 80: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/80.jpg)
Knowledge Based Obligations
Example 2
t = 0
t = 1
t = 2
t = 3
v
m c
t c t c
S ¬S
S
H, 2 |= KuG (t)
2 1
c
c
t c
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 30
![Page 81: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/81.jpg)
Knowledge Based Obligations
Recall that Ann has the (knowledge based) obligation to tell Jillabout her father’s illness (KaG (m)).
Clearly, Ann will not be under any obligation to tell Jill that herfather is ill, if Ann justifiably believes that Jill would not treat herfather even if she knew of his illness.
Thus, to carry out a deduction we will need to assume
Kj(Kusick↔©treat)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 31
![Page 82: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/82.jpg)
Knowledge Based Obligations
Recall that Ann has the (knowledge based) obligation to tell Jillabout her father’s illness (KaG (m)).
Clearly, Ann will not be under any obligation to tell Jill that herfather is ill, if Ann justifiably believes that Jill would not treat herfather even if she knew of his illness.
Thus, to carry out a deduction we will need to assume
Kj(Kusick↔©treat)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 31
![Page 83: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/83.jpg)
Knowledge Based Obligations
Recall that Ann has the (knowledge based) obligation to tell Jillabout her father’s illness (KaG (m)).
Clearly, Ann will not be under any obligation to tell Jill that herfather is ill, if Ann justifiably believes that Jill would not treat herfather even if she knew of his illness.
Thus, to carry out a deduction we will need to assume
Kj(Kusick↔©treat)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 31
![Page 84: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/84.jpg)
Knowledge Based Obligations
A similar assumption is needed to derive that Jill has an obligationto treat Sam.
Obviously, if Jill has a good reason to believe that Ann always liesabout her father being ill, then she is under no obligation to treatSam.
In other words, we need to assume
Kj(msg↔ sick)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 32
![Page 85: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/85.jpg)
Knowledge Based Obligations
A similar assumption is needed to derive that Jill has an obligationto treat Sam.
Obviously, if Jill has a good reason to believe that Ann always liesabout her father being ill, then she is under no obligation to treatSam.
In other words, we need to assume
Kj(msg↔ sick)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 32
![Page 86: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/86.jpg)
Knowledge Based Obligations
A similar assumption is needed to derive that Jill has an obligationto treat Sam.
Obviously, if Jill has a good reason to believe that Ann always liesabout her father being ill, then she is under no obligation to treatSam.
In other words, we need to assume
Kj(msg↔ sick)
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 32
![Page 87: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/87.jpg)
Knowledge Based Obligations
Common Knowledge of Ethicality
These formulas can all be derived for one common assumptionwhich we call Common Knowledge of Ethicality.
1. The agents must (commonly) know the protocol.
2. The agents are all of the same “type” (social utilitymaximizers)
Alternatively, we can argue that Ann has the knowledge basedobligation to send the message because she knows that uponreceiving the message, Uma will change her intentions accordingly(and so, will adopt the intention to treat Sam).
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 33
![Page 88: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/88.jpg)
Knowledge Based Obligations
Common Knowledge of Ethicality
These formulas can all be derived for one common assumptionwhich we call Common Knowledge of Ethicality.
1. The agents must (commonly) know the protocol.
2. The agents are all of the same “type” (social utilitymaximizers)
Alternatively, we can argue that Ann has the knowledge basedobligation to send the message because she knows that uponreceiving the message, Uma will change her intentions accordingly(and so, will adopt the intention to treat Sam).
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 33
![Page 89: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/89.jpg)
Knowledge Based Obligations
Common Knowledge of Ethicality
These formulas can all be derived for one common assumptionwhich we call Common Knowledge of Ethicality.
1. The agents must (commonly) know the protocol.
2. The agents are all of the same “type” (social utilitymaximizers)
Alternatively, we can argue that Ann has the knowledge basedobligation to send the message because she knows that uponreceiving the message, Uma will change her intentions accordingly(and so, will adopt the intention to treat Sam).
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 33
![Page 90: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/90.jpg)
Knowledge Based Obligations
Conclusions
I As you may have noticed, I did not actually present a workingtheory of intention revision! (in progress with Yoav Shoham)
I Pointers to relevant literature left out here are very welcome.
I Many technical questions remain about how to define theoperators B(ϕ : a) and I (a : ϕ), which may fit nicely withJustification Logics.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 34
![Page 91: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/91.jpg)
Knowledge Based Obligations
Conclusions
I As you may have noticed, I did not actually present a workingtheory of intention revision! (in progress with Yoav Shoham)
I Pointers to relevant literature left out here are very welcome.
I Many technical questions remain about how to define theoperators B(ϕ : a) and I (a : ϕ), which may fit nicely withJustification Logics.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 34
![Page 92: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/92.jpg)
Knowledge Based Obligations
Conclusions
I As you may have noticed, I did not actually present a workingtheory of intention revision! (in progress with Yoav Shoham)
I Pointers to relevant literature left out here are very welcome.
I Many technical questions remain about how to define theoperators B(ϕ : a) and I (a : ϕ), which may fit nicely withJustification Logics.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 34
![Page 93: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/93.jpg)
Knowledge Based Obligations
Conclusions
I As you may have noticed, I did not actually present a workingtheory of intention revision! (in progress with Yoav Shoham)
I Pointers to relevant literature left out here are very welcome.
I Many technical questions remain about how to define theoperators B(ϕ : a) and I (a : ϕ), which may fit nicely withJustification Logics.
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 34
![Page 94: Towards a Theory of Intention Revisionrobotics.stanford.edu/~epacuit/talks/wkshp-dynintention.pdf · Towards a Theory of Intention Revision Workshop on the Dynamics of Intention and](https://reader033.vdocument.in/reader033/viewer/2022053010/5f0dde6c7e708231d43c7c0d/html5/thumbnails/94.jpg)
Knowledge Based Obligations
Thank You!
Eric Pacuit: Towards a Theory of Intention Revision, Workshop on the Dynamics of Intention and Preference 35