how to get paidgroups.di.unipi.it/~morge/publis/morge08emse_show.pdf · [morge, icc’07] [morge et...

How to get PAID ?June 2008.

Maxime Morge

Università di Pisa

Practical Applications of Argumentation and Multiagent Systemsfor Semantics Integration in Distributed Environment

June 2008. wwww.argugrid.eu

Introduction

Motivations and challenges

Requirement for application softwares in a global, complex, interconnected,and dynamic environments (services, industries, . . . )

Reusabilty, Interoperability, Flexibility, Adaptation.

Efficiency, Autonomy, Intelligence, Scalability.

Reliability, Fault-tolerance, Robustness, Security, Trust.

Argumentation support

DIAL

Debate support

[Morge, ICC’07] [Morge et al, EUMAS’07] [Morge et al, AISB’07]


Introduction

Conflicts resolution for muliagent programming

Our multiagent systems are composed of proactive agents (eco./logic.)within distributed environments (social, physical) interacting eachother (with ACL) in accordance with norms (contracts).

Conflicts arise in:• ontologies i.e.concepts hierarchies;

• agent state-of-mind, i.e.beliefs, goals, decisions;

• agent’s mind, i.e.the behaviour;

• coordination,i.e.the negotiation.

Conflicts resolution:• heterogeneous descriptions[Morge et al, AO07]

• epistemic/practical reasoning[Morge, AIL06],[Morge ArgMAS 07]

• dialectical agent architecture[Morge et al, JFSMA’08]

• dispute resolution [Morge, AIL06][Morge et al, EUMAS’07]


Introduction

Outline

Introduction

ARGUGRID

Agent reasoning [Morge ArgMAS 07]

Negotiation [Morge et al, EUMAS’07]

Conclusion and future works


ARGUGRID

ARGUmentation as a foundation for the semantic GRID

Provide a new model for argumentative agents populating andevolving within a trusted grid.Provide a new model for the specification, creation, operation anddissolution of Virtual Organizations over the grid usingargumentation.Design an architecture for the semantic grid to support argumentativeagents and VOs.Develop a grid-based platform to support the implementation ofmodels and architecture and assess the approach.Experiment with and evaluate the models, architecture and platform inthe context of concrete applications for e-business.


ARGUGRID

Global Picture of the ARGUGRID platform


ARGUGRID

EO Scenario: oil spill detection case [ARGUGRID D1.2]

��

��

��

��

��

��

��

��

VO

Client #1

Provider #1

Provider #3 Provider #4

Provider #2

Images Images

Postprods Postprods


Agent reasoning

Outline

Introduction

ARGUGRID





Agent reasoning Semantics

Assumption-based argumentation for multi-criteria

decision making [Morge ArgMAS 07]

A B

C D

Which contract ?

The service d provided by Provider #1.

Why not the service c?

Since it’s too expensive.

Why not Provider #2 ?

Since Provider #2 will overcharge the price.

Decision analysis (influence diagrams).

Logic language (goals/decisions/beliefs).

User’s preferences (priorities/utilities/uncertainty).

Suggested decisions (credulous semantics).

Procedure implementation (http://margo.sourceforge.net).


Agent reasoning Decision Analysis

A model of multi-criteria decision problems with

incomplete knowledge

good_deal

supplier provision

rcost rqos cost qos

Proposal(x , y)

price(y)resolution(y)will(x , do)

Test(tid , x , done)

type(tid , tp)

int(tid , i)

time(tid , t)


Agent reasoning Decision framework

Knowledge, Goals, Decisions, and Priority

A decision framework is a tuple D = 〈L,Asm, I , T , P 〉, where:

L is the language which captures the state-of-mind(goals/decisions/beliefs);

Asm, is a set of literals which are taken for granted if there is noevidence to the contrary, called assumptions;

I is the incompatibility relation, i.e. a binary relation whichcaptures the conflicts between the statements;

T is the theory which gathers goal/decision/epistemic rules;

P ⊆ T ×T is a preorder, called the priority relation, which captures

the uncertainty of beliefs,

the priority amongst goals,

and the expected utilities of the decisions.


Agent reasoning Decision framework

Goal rules, decision rules, and epistemic rules

T

r1(tid): will(x , do)← Test(tid , x , do), type(tid , neutral), int(tid , i), time(tid , t)r2(tid): will(x , do)← Test(tid , x , do), type(tid , tp), int(tid , direct), time(tid , t)r′1(tid): will(x , do)← Test(tid , x , do), type(tid , concurrent), int(tid , i), time(tid , t)r′2(tid): will(x , do)← Test(tid , x , do), type(tid , tp), int(tid , observed), time(tid , t)f1: Test(Observer #11, Provider #2,¬overcharge)f2: Test(Observer #11, Provider #2,¬underquality)f3: type(Observer #11, neutral)f4: time(Observer #11, 1)f5: Test(Client #21, Provider #2, overcharge)f6: Test(Client #21, Provider #2, underquality)f7: type(Client #21, concurrent)f8: time(Client #21, 1)f9: Test(Client #11, Provider #1, overcharge)f10: Test(Client #11, Provider #1,¬underquality)f11: type(Client #11, neutral)f12: int(Client #11, direct)f13: time(Client #11, 3)f14: Test(Observer #12, Provider #1,¬overcharge)f15: Test(Observer #12, Provider #1, underquality)f16: type(Observer #12, neutral)f17: time(Observer #12, 2)f18: price(d, high)f19: resolution(d, low)f20: price(c, low)f21: resolution(c, high)f22: price(e, low)f23: resolution(e, low)f24: price(f, high)f25: resolution(f, high)

T

r012: good_deal← supplier, provision

r134: supplier← rcost, rqos

r256: provision← cost, qos

r01: good_deal← supplier

r13: supplier← rcost

r25: provision← cost

r02: good_deal← provision

r14: supplier← rqos

r26: provision← qos

T

r31 : rcost← Proposal(x , y), will(x ,¬overcharge)r41 : rqos← Proposal(x , y), will(x ,¬underquality)r51 : cost← Proposal(x , y), price(y , low)r62 : qos← Proposal(x , y), resolution(y , high)r52 : cost← Proposal(x , y), price(y , high)r61 : qos← Proposal(x , y), resolution(y , low)


Agent reasoning Arguments

Rebutting and undermining attacks

Rebutting attack between conflicting conclusions:

The service d is a good deal because it is cheap;

The service c is not a good deal because it is a mess.

good_deal ¬good_deal

S(Provider #2, c)

cheap

S(Provider #2, c)

unreliable



Rebutting and undermining attacks

Undermining attack toward non-provable assumptions:

The service c is a good deal because it is cheap and

it is not provable that it is a mess;

The service c is a mess.

unreliablegood_deal

S(Provider #2, c) not unreliable S(Provider #2, c)

June 2008. Page 13 wwww.argugrid.eu


The strengths of arguments

Qualitative or quantitative preferences, s.a.:

the likelihood of beliefs.the priority amongst goals.the expected utililies of decisions.

good_deal

cheap

S(Provider #2, d)

¬good_deal

unreliable

S(Provider #2, c)

≤

Defeat


Agent reasoning Proof procedure

Dialectical enquiry of MARGO

A




A B




A B C




A B C

A (procedural) dialectical proof procedure:

⇔ declarative model-theoretic semantic (soundness/completeness);

interleaves arguments construction/dispute (efficiency);

with game rule (turn-taking, burden of proof, backtraking);

preference relaxation, assumptions (strategy).


Negotiation

Outline

Introduction

ARGUGRID





Negotiation Agent Architecture

A Modular Architecture of Service Oriented Agents

IMQ OMQ IEQ OEQ

IKB ALSKB PL

CM

IDMM SDMM SIM

GUI

PIM

Module

Other agents andregistry

Interaction with externalities

Module reads data

Module reads/writes data

Data


Negotiation VO formation

Life-cycle of VOs [ARGUGRID D3.2]

phase sub-phases tasks implementation

Formation Initiation goal identification GUI, CM, IDMMpartner discovery SDMM, PIM, CM

Configuration partner selection SDMMestablish roles SDMM, SIMnegotiation SDMM, SIM, CM

Operation execution KDE, GRIAreformation MAS

reportage MAS, GRIAmonitoring GRIA

Dissolution reportage GRIAevaluation IDMM, SDMM, MAS


Negotiation Social Interaction

Dialogue for conflicts explanation

Client #1 Provider #2

question(good_deal, S(y), ∅)

assert(good_deal, S(d), ∅)

assert(good_deal, S(c), ∅)

why(good_deal, S(c), ∅)

assert(cost, S(c), price(c , 20))

assert(cost, S(e), price(c , 20))

accept(good_deal, S(e), ∅)



Negotiation protocol for the requester

evaluatecontract

challengeevaluate

send question

send assert

receive assert

receive assert

receive assert

receive whyreceive accept

send assert

send accept

send why

receive withdraw

send withdraw



A LCC Protocol for Negotiation of services

Lightweight Coordination Calculus:

boot strap mechanism protocol/role/participants;

pre/post conditions to prompt/update the decison/commitments;

write once-execute everywhere.

a(requester(G,D,K,Ag2),AID)::=question(G , D, K) ⇒ a(provider(G , D, K), Ag2) then

assert(G , D, K) ⇐ a(provider(G , D, K), Ag2) then

a(evaluator(G , D, K), AID).

a(evaluator(G,D,K,Ag2),AID)::=accept(G , D, K) ⇒ a(interlocutor(G , D, K), Ag2)← acceptable(G , D, K)or

(assert(G , D2, K) ⇒ a(interlocutor(G , D, K), Ag2)← counter(G , D, K , D2) then

a(interlocutor(G , D2, K), AID))or

(why(G , D, K) ⇒ a(interlocutor(G , D, K), Ag2)← question(G , D, K) then

(assert(G2, D, K2) ⇐ a(interlocutor(G , D, K), Ag2) then

a(evaluator(G2, D, K2), AID)) or

withdraw ⇐ a(interlocutor(G , D, K), Ag2)).

a(provider(G,D,K),Ag2)::=question(G , D, K) ⇐ a(requestor(G , D, K , Ag2), AID) then

assert(G , D, K) ⇒ a(requestor(G , D, K , Ag2), AID)← offer(G , D, K) then

a(interlocutor(G , D, K), Ag2).

a(interlocutor(G,D,K),Ag2)::=accepted(G , D, K) ←accept(G , D, K) ⇐ a(evaluator(G , D, K , Ag2), AID)or

(assert(G , D2, K) ⇐ a(evaluator(G , D, K , Ag2), AID) then

a(evaluator(G , D2, K), Ag2))or

(why(G , D, K) ⇐ a(evaluator(G , D, K , Ag2), AID) then

(assert(G2, D, K2) ⇒ a(evaluator(G , D, K , Ag2), AID)← counter(G , D, K , G2, K2) then

a(interlocutor(G2, D, K2), Ag2)) or

withdraw ⇒ a(evaluator(G , D, K , Ag2), AID)).


Negotiation Bilateral concession

Acceptability space of participants

After M3 At the end

bSb(d)

b

Sb(c)

priceb

resolutionb

Client #1

Provider #2bSb(d)×Sb(d)

×

Sb(c)bcb

Sb(e)

priceb

resolutionb

Client #1

Provider #2


Negotiation System deployment

GOLEM Containers [ARGUGRID D3.3]

Client #1 Client #2

Registery

Reg

iste

ryC

onne

ctor

RegisteryConnector

Reg

iste

ryC

onne

ctor

container

agent communicationor agent/service discovery

Connector

Provider #1 Provider #2

Provider #3

Provider #4


Summary

Outline

Introduction

ARGUGRID





Summary

Dialectics for semantic integration in distributed

environments

Summary:• Arguing for decison under uncertaintyconflicts over agent’s statements[Morge ArgMAS 07]

• Dialectical agent architectureconflicts over agent’s motivations[Morge et al, JFSMA’08]

• Dispute for service compositionconflict over agents’ interests[Morge et al, EUMAS’07]

Future works:• Arguing for planningwith normative/trustreasoning.

• Multi-thread prototypewith experimental validation

• Game theoryfor interactive strategy


Summary

References

Morge, M.Collective Decision making process to compose divergent interests and perspectives.

Special issue on Argumentation in Journal of Artificial Intelligence and Law, p 75-92,Springer-Verlag.

Morge, M. and Routier, JCDebating over heterogeneous descriptions.

Applied Ontology, Special issue on Formal Ontology for Communicating Agents, Volume 2,Number 3-4, pages 333-349, IOS press.

Maxime Morge.The hedgehog and the fox. An argumentation-based decision support system.

In Rahwan, I., Parsons, S., Reed, C., Argumentation in Multi-Agent Systems: FourthInternational Workshop ArgMAS, Revised Selected and Invited Papers, vol 4946 of LectureNotes in Artificial Intelligence, p 114–131, Springer-Verlag.

Maxime MorgeSe concerter à l’aide d’un système multi-agents.

Humanités numériques 2. Socio-informatique et démocratie cognitive, Cognition ettraitement de l’information, p 117-126, Hermès science publication.


Summary

References (bis)

M. Morge, J. McGinnis, S. Bromuri, F. Toni, P. Mancarella, K. Stathis.Toward a Modular Architecture of Argumentative Agents to Compose Services.

Proc. of EUMAS’07, Hammamet, Tunisia 2007

Morge, M and Mancarella, P.Argumentation-based decision making for selecting communication services in

ambient home environments.

Proc. of AISB Convention, Newcastle University, UK, April, 2007.

M. Morge, K. Stathis and L. VercouterArguing over motivations for self-adaptation within the V3A-architecture.

Proc. of 16ème Journées Francophones sur les Systèmes Multi-Agents (JFSMA’08), Brest,France 2008.

T. Stournaras, editor.eBusiness application scenarios.Deliverable document D1.2 ARGUGRID, 2007.

Phan Minh DungOn the acceptability of arguments and its fundamental role in nonmonotonic

reasoning, logic programming and n-person games.

Artificial Intelligence, Vol. 77, Issue 2, p. 321-357, 1995


Summary

References (bis)

J. McGinnis, editor.Virtual Organisations as Agent Societies: Phases.Deliverable document D3.2 ARGUGRID, 2007.

S. Bromuri, editor.GOLEM.Deliverable document D3.3 ARGUGRID, 2007.


Summary Arguing over motivations

The V3A Architecture [Morge et al, JFSMA’08]

AS

Body

persKBenv K

Bint

KBorgK

Bagt

KB

env1

KBenv2

KBuser

f1

f2

f3

f4 f5

f6

f7

KB

int1

KBint2

Component-basedarchitecture

Vowels approach

Dialogue amongst facets

Personality-basedresolution of conflicts

High-level control of theautonomy and thebehaviour

⇒ Self-adaptation in openMAS


how to get paidgroups.di.unipi.it/~morge/publis/morge08emse_show.pdf · [morge, icc’07] [morge et...

Documents