Upload
federico-cerutti
View
198
Download
0
Embed Size (px)
Citation preview
Argumentation-Based Decision Making andStructural Models of Personality
Pietro Baroni and Federico Cerutti and MassimilianoGiacomin and Giovanni Guida
Dipartimento di Ingegneria dell’Informazione, Universita di BresciaVia Branze 38, I-25123 Brescia, Italy
December 3, 2010
ARGAIP 2010
c© 2010 Federico Cerutti <[email protected]>
“Good Advice”
The advice should be presented in a form which can be readilyunderstood by decision makers
There should be ready access to both information and reasoningunderpinning the advice
If decision support involves details which are unusual to thedecision maker, it is of primary importance that s/he can discussthese details with his advisor [Girle et al., 2003]
c© 2010 Federico Cerutti <[email protected]> December 2010 2
Practical Reasoning about “What to do”
Knowledge Representation Computation of Outcomes
Knowledge RepresentationComputation of Outcomes
Conclusions and Future works
c© 2010 Federico Cerutti <[email protected]> December 2010 3
Argument (and Attack) schemes
Argument schemes for knowledge representation [Walton, 1996]
Structure which contains the information supporting a givenconclusion
Modelling conflicts by “attack scheme”
Structure which contains the information supporting a givenconflict
c© 2010 Federico Cerutti <[email protected]> December 2010 4
The main concepts
Circumstance: a state of the world
Fact: a particular circumstance assumed to be true
Goal: a state of the world we want to achieve
Action: support for the achievement of a goal
Emotion: “a strong feeling deriving from one’s circumstances,[or] mood[. . . ]”
Preference: “[. . . ] a greater liking for one alternative overanother or others [. . . ]”
Value: “Worth or worthiness [. . . ] in respect of rank or personalqualities”
Must Value: a value that we are committed to promote
c© 2010 Federico Cerutti <[email protected]> December 2010 5
A scenario (1)
I have to go to theAmazon Rainforest
I want to help people!
c© 2010 Federico Cerutti <[email protected]> December 2010 6
A scenario (2)
Ough! My Hernia! Doc! What canI do?
You can:1) Do a discectomy surgery2) Have a long non-invasive treatment3) Take analgesics
Mmmmhh:1) I promote my safety2) I promote my safety3) I promote charity because I can go to Brazil
Let us assume that these actions are mutually exclusive
c© 2010 Federico Cerutti <[email protected]> December 2010 7
A scenario (2): practical arguments
A2: circumstance: in my situation,action: discectomy surgery,goal: reducing disc herniation,value: safety,sign: +.
[Greenwood et al., 2003]
c© 2010 Federico Cerutti <[email protected]> December 2010 8
Mumble
A scenario (2): practical arguments
α [PAtS1]: source: A1,target: A2,conditions: A1.action and A2.actionare incompatible.
c© 2010 Federico Cerutti <[email protected]> December 2010 8
Mumble
A scenario (2): values arguments
V1: value: charity.V2: value: safety.
They are “socially accepted values”.
c© 2010 Federico Cerutti <[email protected]> December 2010 9
Mumble
A scenario (2): values defence (i)
β [VDes1]: source: V2,target: α,conditions: α.target.value = V2.value andα.source.value 6= A2.value.
Each value is committed to protect each practicalargument that promotes it! And. . .
c© 2010 Federico Cerutti <[email protected]> December 2010 10
Mumble
A scenario (2): values defence (ii)
γ [VDes2]: source: V1,target: β,conditions: β.source 6= V1 andβ.target.source.value 6= V1.value.
Each value is committed to protect each attackgrounded on each practical argument that promotes it!
c© 2010 Federico Cerutti <[email protected]> December 2010 11
Mumble
A scenario (2): values defence (ii)
β [VDefence]: defending: A2,defended: V2,conditions: defending.value =defended.value and defended.sign = +.
c© 2010 Federico Cerutti <[email protected]> December 2010 11
Mumble
A scenario (2): values defence (ii)
γ [VDefence]: defending: A1,defended: V1,conditions: defending.value =defended.value and defended.sign = +.
c© 2010 Federico Cerutti <[email protected]> December 2010 11
Mumble
A scenario (3)
But I can't do a discetomysurgery since I have a historyof anaesthesia allergy
c© 2010 Federico Cerutti <[email protected]> December 2010 12
A scenario (3): contradictory pract. arg. (i)
A4: circumstance: alternative anaesthesia not availablefor discectomy,action: not discectomy surgery,goal: risk of shock,value: safety,sign: −.
c© 2010 Federico Cerutti <[email protected]> December 2010 13
Mumble
A scenario (3): contradictory pract. arg. (i)
δ [PAtS2]: source: A4,target: A2,conditions: A4.action = ¬ A2.action.
c© 2010 Federico Cerutti <[email protected]> December 2010 13
Mumble
A scenario (3): contradictory pract. arg. (ii)
ε [VAAtS]: source: A4,target: β,conditions: A4.action = ¬ β.defended.actionand A4.value = β.defended.value andA4.sign = − and β.defended.sign = +.
c© 2010 Federico Cerutti <[email protected]> December 2010 14
Mumble
A scenario (4)
But I can't do a discetomysurgery since I have a historyof anaesthesia allergy
Don't worry... There is a kindof anaesthesia you are not allergic to
c© 2010 Federico Cerutti <[email protected]> December 2010 15
A scenario (4): factual arguments
A5: circumstances: alternative anaesthesia is availablefor discectomy.
c© 2010 Federico Cerutti <[email protected]> December 2010 16
Mumble
A scenario (4): factual arguments
ι [FAtS1]: source: A5,target: A4,conditions: A5.circumstances =¬ A4.circumstances.
c© 2010 Federico Cerutti <[email protected]> December 2010 16
Mumble
A scenario (5)
OK. With surgery I will getover my illness in very fewdays. But I'm frightened by surgery...
Noo... I categorically rejectthe idea!
c© 2010 Federico Cerutti <[email protected]> December 2010 17
A scenario (5): preferences. . .
P1: preferred: A2,notpreferred: A3.
c© 2010 Federico Cerutti <[email protected]> December 2010 18
Mumble
A scenario (5): preferences. . .
λ [PRAtS]: source: P1,target: κ,conditions: P1.preferred = κ.targetand P1.notpreferred = κ.source.
c© 2010 Federico Cerutti <[email protected]> December 2010 18
Mumble
A scenario (5): . . . and emotions
E1: emotion: surgery frightening.
Are emotions rational?
c© 2010 Federico Cerutti <[email protected]> December 2010 19
Mumble
A scenario (5): . . . and emotions
µ [EAtS1]: source: E1,target: A2,conditions: e(E1,A2) = −.
with e(E1,A2) = −
c© 2010 Federico Cerutti <[email protected]> December 2010 19
Mumble
A scenario (6)
OK. I must care about my health.Then, I will help people again!
c© 2010 Federico Cerutti <[email protected]> December 2010 20
A scenario (6): Must Value
MUST V2: value: safety.
This is again a rational choice!
c© 2010 Federico Cerutti <[email protected]> December 2010 21
Mumble
A scenario (6): Must Value
ν [MAtS2]: source: MUST V2,target: γ,conditions: MUST V2.value =γ.target.source.value andMUST V2.value 6= γ.source.value.
c© 2010 Federico Cerutti <[email protected]> December 2010 21
Mumble
Why three boxes?
c© 2010 Federico Cerutti <[email protected]> December 2010 22
Why three boxes?
1 The set of arguments that take into accountvalues;VAS arguments.
2 The set of arguments whose bases are not rational;EAS arguments.
3 The set of arguments that deal with both facts and values;FAS, PAS, PRAS, and MAS.
c© 2010 Federico Cerutti <[email protected]> December 2010 22
Look familiar?
c© 2010 Federico Cerutti <[email protected]> December 2010 23
Look familiar?
1 The Superego strives to act in a socially appropriate manner.
2 The Id is the innate part of our personality and is based on thepleasure principle.
3 The Ego represents what may becalled reason and common senseand has to conciliate the innateinstincts and the social constraints.
c© 2010 Federico Cerutti <[email protected]> December 2010 23
Knowledge Representation
Computation of OutcomesConclusions and Future works
c© 2010 Federico Cerutti <[email protected]> December 2010 24
Argumentation Framework for Decision SupportProblemAn Argumentation Framework for Decision Support Problem(AFDSP ) is a 12-ple 〈APAS , APRAS , AEAS , AV AS , AFAS ,AMAS ,RPAS , RPRAS , REAS , RV AS , RFAS , RMAS〉 s.t.:
APAS is a set of instances of PAS;
APRAS is a set of instances of PRAS;
AEAS is the set of instances of EAS;
AV AS is a set of instances of VAS;
AFAS is a set of instances of FAS;
AMAS is a set of instances of MAS;
RPAS is a set of instances of PAtS1 and PAtS2;
RPRAS is a set of instances of PRAtS;
REAS is a set of instances of EAtS1 and EDefence;
RV AS is a set of instances of VAtS, VDeS1, VDeS2, VAAtS;
RFAS is a set of instances of FAtS;
RMAS is a set of instances of MAtS1, and MAtS2.
c© 2010 Federico Cerutti <[email protected]> December 2010 25
Argumentation Framework with RecursiveAttacks (AFRA)
An Argumentation Framework with Recursive Attacks (AFRA) is apair 〈A,R〉 where A is a set of arguments and R is a set of attacks,namely pairs (A,X ) s.t. A ∈ A and (X ∈ R or X ∈ A).
Given an attack α = (A,X ) ∈ R, we will say that A is the source ofα, denoted as src(α) = A and X is the target of α, denoted astrg(α) = X . [Baroni et al., 2009b]
c© 2010 Federico Cerutti <[email protected]> December 2010 26
From AFDSP to AFRA
The instances of argument schemes in AFDSP compose the set ofarguments in AFRA and the instances of attack schemes in AFDSPgive rise to the attack relation in AFRA.
Let Φ = 〈APAS , APRAS , AEAS , AV AS , AFAS , AMAS ,RPAS , RPRAS ,REAS , RV AS , RFAS , RMAS〉 be an AFDSP , the correspondingAFRA is defined as Γ = 〈A,R〉 s.t.
A = APAS ∪ APRAS ∪ AEAS ∪ AV AS ∪ AFAS ∪ AMAS ; and
R = RPAS ∪RPRAS ∪REAS ∪RV AS ∪RFAS ∪RMAS .
c© 2010 Federico Cerutti <[email protected]> December 2010 27
Defeat relation
Definition (Direct Defeat)
Let 〈A,R〉 be an AFRA, V ∈ R, W ∈ A∪R, then V directly defeatsW iff W = trg(V ).
Definition (Indirect Defeat)
Let 〈A,R〉 be an AFRA, V ∈ R, W ∈ A, if V directly defeats Wthen ∀α ∈ R s.t. src(α) = W , V indirectly defeats α.
Definition (Defeat)
Let 〈A,R〉 be an AFRA, V ∈ R, W ∈ A ∪R, then V defeats W ,denoted as V →R W , iff V directly or indirectly defeats W .
c© 2010 Federico Cerutti <[email protected]> December 2010 28
Semantics notions
Definition (Conflict–free)
Let 〈A,R〉 be an AFRA, S ⊆ A∪R is conflict–free iff @V ,W ∈ S s.t.V →R W .
Definition (Acceptability)
Let 〈A,R〉 be an AFRA, S ⊆ A ∪R, W ∈ A ∪R, W is acceptablew.r.t. S iff ∀Z ∈ R s.t. Z →R W ∃V ∈ S s.t. V →R Z .
Definition (Admissible set)
Let 〈A,R〉 be an AFRA, S ⊆ A∪R is admissible iff it is conflict–freeand each element of S is acceptable w.r.t. S.
c© 2010 Federico Cerutti <[email protected]> December 2010 29
Preferred Extension
A preferred extension is a maximal (w.r.t. set inclusion) admissibleset.
c© 2010 Federico Cerutti <[email protected]> December 2010 30
Recalling the example. . .
A1
A2 A3A4
V2
V1
d
a
b
g
e
A5
iP1
E1
k l
m MUSTV2
n
Ego
Superego
Id
c© 2010 Federico Cerutti <[email protected]> December 2010 31
. . . and the preferred extension
A3
V2
V1
b
A5
iP1
E1
l
m MUSTV2
n
Ego
Superego
Id
c© 2010 Federico Cerutti <[email protected]> December 2010 32
Knowledge Representation
Computation of Outcomes
Conclusions and Futureworks
c© 2010 Federico Cerutti <[email protected]> December 2010 33
Conclusions
Extension of a previous proposal [Baroni et al., 2009a]
Human emotions in decision making process
Main contribution: preliminary description of a mapping betweenargumentation based decision process and Freud’s three entitiesmodel of personality
c© 2010 Federico Cerutti <[email protected]> December 2010 34
Future works
Fill the gap between attack schemes and critical questions
Deep analysis on Superego and Id:
Enhancing the relevant argument schemesMore articulated personality models
c© 2010 Federico Cerutti <[email protected]> December 2010 35
Thank you
c© 2010 Federico Cerutti <[email protected]>
References
[Baroni et al., 2009a] Baroni, P., Cerutti, F., Giacomin, M., andGuida, G. (2009a).An argumentation-based approach to modeling decision supportcontexts with what-if capabilities.In AAAI Fall Symposium. Technical Report SS-09-06, pages 2–7.AAAI Press.
[Baroni et al., 2009b] Baroni, P., Cerutti, F., Giacomin, M., andGuida, G. (2009b).Encompassing attacks to attacks in abstract argumentationframeworks.In Proc. of ECSQARU 2009, pages 83–94, Verona, I.
[Girle et al., 2003] Girle, R., Hitchcock, D. L., McBurney, P., andVerheij, B. (2003).Decision support for practical reasoning: A theoretical andcomputational perspective.In Reed, C. and Norman, T. J., editors, Argumentation Machines.New Frontiers in Argument and Computation, pages 55–84. Kluwer.
[Greenwood et al., 2003] Greenwood, K., Capon, T. B., andMcBurney, P. (2003).Towards a computational account of persuasion in law.In Proc. of ICAIL ’03, pages 22–31.
[Walton, 1996] Walton, D. N. (1996).Argumentation Schemes for Presumptive Reasoning.Lawrence Erlbaum Associates.
c© 2010 Federico Cerutti <[email protected]> December 2010 37