Cerutti--AAAI Fall Symposia 2009

University of Brescia
Dipartimento di Elettronica per l'Automazione
Knowledge Engineering and Human-Computer Interaction Research Group

An Argumentation-based
Approach to Modelling Decision
Support Contexts with What-If
Capabilities
Pietro Baroni, Federico Cerutti, Massimiliano Giacomin and Giovanni Guida

AAAI 2009 Fall Symposium Series
The Uses of Computational Argumentation
Arlington, November, 5, 2009

© 2009 Federico Cerutti <federico.cerutti@ing.unibs.it>

“Good advice”
 The advice should be presented in a form which can
be readily understood by decision makers
 There should be ready access to both information and
reasoning underpinning the advice
 If decision support involves details which are
unusual to the decision maker, it is of primary
importance that s/he can discuss these details with
his advisor

Girle et al., 2003
Slide 2 © 2009 Federico Cerutti <federico.cerutti@ing.unibs.it>

Transparency about the advice

Practical reasoning
about “what to do”

Knowledge Computation of
representation outcomes


Knowledge representation
➢Computation of outcomes
➢ Conclusions and future works


Argument (and attack) schemes
 Use of argument scheme to represent the knowledge
 Structure which contains the information in favour of
a given conclusion

 Introduction of a possible modelling of conflicts by
“attack scheme”
 Structure which contains the information in favour of
a given conflict


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
 Preference: “[…] a greater liking for one alternative
over another or others […]”
 Value: “Worth or worthiness […] in respect of rank
or personal qualities”
 Must Value: a value that we commit to promote

Example (1)
 Treatment for a patient threatened by blood clotting
 Goal: obtaining a low platelet adhesion
 The available knowledge base concerning
treatments:
 Administer Aspirin (value of Safety)
 Administer Chlopidogrel (value of Safety)
 Do nothing (value of Cost)


Formal counterpart (1)
Practical Args (from Atkinson et al.)

PAS: A2
Circumstances: given patient's situation
Action: we should administer aspirin
Goal: reducing blood clotting
Value: Safety
Sign: +

The Attacks among PAS

PAtS1: α
Source: A1
Target: A2
Conditions:
A1.action and A2.action
are incompatible


The Values

VAS: V1
Value: Cost


The Defences from the Values

VDeS1: β
Source: V2
Target: α
Conditions:
α.target.value = V2.value,
α.source.value ≠ V2.value

PAtS1: α
…


VDeS2: γ
Source: V1
Target: β
Conditions:
β.source ≠ V1,
β.target.source.value ≠ V1.value

VDeS1: β
Source: V2
Target: α
Conditions:

PAtS1: α
…


VDefence: γ VDeS2: γ
Defending: A1 Source: V1
Defended: V1 Target: β
Conditions:
β.source ≠ V1,
VDefence: β
β.target.source.value ≠ V1.value
Defending: A2
Defended: V2
VDeS1: β
Source: V2
Target: α
Conditions:

PAtS1: α
…


Example (2)
 From patient's file we learn that he has a history of
gastritis
 Then we should not administer Aspirin without a
proton pump inhibitor
 In fact, it gives rise to risk of ulceration
 And it will demote the value of Safety
 As far as we know, no proton pump inhibitor is
available


A PAS with negative sign...

PAS: A4
Circumstances: proton pump unavailable
Action: we should not administer aspirin
Goal: risk of ulceration
Value: Safety
Sign: -

...and the relative attacks

PAtS2: δ
Source: A4
Target: A2
Conditions:
A4.action = ¬ A2.action


...and the relative attacks
VAAtS: ε
Source: A4
Target: β
Conditions: A4.circumstance =
β.defended.circumstance, A4.action =
¬ β.defended.action, A4.goal =
β.defended.goal, A4.value =
β.defended.value, A4.sign = -,
β.defended.sign = +

PAtS2: δ
Source: A4
Target: A2
Conditions:
A4.action = ¬ A2.action


Example (3)
 Suppose now that the assumption that no proton
pump inhibitor is available reveals to be false
 Suppose also that between aspirin and chlopidogrel a
doctor prefers to administer aspirin because it is in
stock and immediately available


A Fact

FAS: A5
Circumstances: a
proton pump is
available


An “undercut”

FAtS: ζ
Source: A5
Target: A4
Conditions: A5.circumstances=
¬ A4.circumstances

A preference

PRAS: P1
Preferred: A2
Notpreferred: A3


The greater liking
PAtS: η
...

FAtS: θ
Source: P1
Target: η
Conditions: P1.preferred =
η.target, P1.notpreferred = η.source

Example (4)
 Determine the ultimate decision outcome
 Achieve the goal of reducing blood clotting
 It promotes the value of Safety
 We must promote the value of Safety


The Must Value
MAS: MV2
Value: Safety


What-If Scenario
MAtS2: κ
Source: MV2
Target: γ
Conditions: MV2.value = γ.target.source.value,
MV2.value ≠ γ.source.value

VDeS2: γ
...


Small Summary
Argument Scheme Taxonomy
 Practical Argument Scheme
 Factual Argument Scheme
 Value Argument Scheme
 Preference Argument Scheme
 Must Argument Scheme


In the example


Small Summary
Attack Scheme Taxonomy (1)
 PAtS1: incompatible actions
 PAtS2: “rebuttal”
 VAtS : incompatible values
 VDefence (VDeS[1-2]): a value protects both the
arguments which promote it and the attacks sourced
from that arguments
 VAAtS: if a practical argument P suggests not to
perform an action A since it demotes a value V, if P
will be considered acceptable, then V cannot defend
the argument whose action is A

In the example (1)


Small Summary
Attack Scheme Taxonomy (2)
 FAtS: “undercut”
 PRAtS: someone told us that an attack cannot be
considered since an external preference
 MAtS1: an instance of Must Argument Scheme has
to protect the related Value argument against the
incompatible values
 MAtS2: an instance of Must Argument Scheme has
to protect the instances of VDefence which start from
the related Value argument


In the example (2)


➢ Knowledge representation
Computation of outcomes
➢ Conclusions and future works


Argumentation Framework for
Decision Support Problem


Argumentation Framework with
Recoursive Attacks (AFRA)


From AFDSP to AFRA


AFRA: Defeat relation


AFRA: Admissibility


AFRA: Preferred Extension


Recalling the example...


...and the preferred extension


➢ Knowledge representation
➢ Computation of outcomes

Conclusions and
future works


Conclusions
 Preliminary investigation about formalisation of
decision support problems
 Three main contributions:
 The role of attack schemes
 Attacks to attacks in practice
 Support to “What-if” reasoning


Future works
 Knowledge representation:
 Enhancing attacks schemes
 Ontological status of attacks
 Multiple What-if Situations
 Computation of outcomes
 Further investigation on the theoretical bases of AFRA
 Argumentation semantics in this context


Open questions
 The notion of attack scheme: soundness and
usefulness
 Attack schemes and critical questions
 What-if only w.r.t. Values


Cerutti--AAAI Fall Symposia 2009

More Related Content

Viewers also liked

Similar to Cerutti--AAAI Fall Symposia 2009

More from Federico Cerutti

Cerutti--AAAI Fall Symposia 2009