[PPT] - Formalising Situatedness and Adaptation in Electronic Institutions PowerPoint Presentation

SLIDE 1

Formalising Situatedness and Adaptation in Electronic Institutions

Jordi Campos1, Maite López-Sánchez1,

J. A. Rodríguez-Aguilar2, Marc Esteva2

1 Universitat de Barcelona (UB) 2 Institut d'Intel·ligència Artificial (IIIA-CSIC)

SLIDE 2

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 2/27

Introduction: Context

MAS context:

– Organisations in MAS

approaches:

– from a Participant perspective – from an Organisation perspective

based on:

– social conventions (include norms)

Electronic Institution (EI):
normative MAS framework

MAS Organisations Participant perspective Organisation perspective

SLIDE 4

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 4/27

Introduction: Motivation

– Motivation:

extending EIs to other contexts:

– use EI's regulation capabilities

to enhance an existing MAS.

– adapt EI's social conventions to a

dynamic set of participant agents.

– Problem:

EI needs to be aware of world
EI should adapt

– Our approach:

To bring to an organisational

level two agent properties:

Existing dynamic social system (world) Electronic Institution (EI) Agent1 Agent2 interaction to enhance

situatedness (SEI)
adaptation (AEI)

Rules => Rules' (adaptive)

SLIDE 5

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 5/27

Electronic Institution

Agent1 Agent2 EI External agents

– EI = < DF, DC >

DF: Dialogical Framework
DC: Deontological Comp.

DF DC

SLIDE 7

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 7/27

Electronic Institution: DF

Agent1 Agent2 External agents inform(a1:role1, a2:role2, contents)

– EI = < DF, DC >

Dialogical Framework

– Ontology – Information Models – Social Structure – Communication Lang.

Deontological Comp.

DF DC EI role1 attribs1 role2 attribs2

SLIDE 8

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 8/27

Electronic Institution: PS

– EI = < DF, DC >

Dialogical Framework

– Ontology – Information Models – Social Structure – Communication Lang.

Deontological Comp.

– Performative Structure – Normative Structure

Agent1 Agent2 External agents EI DF DC

SLIDE 9

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 9/27

Electronic Institution: NS

– EI = < DF, DC >

Dialogical Framework

– Ontology – Information Models – Social Structure – Communication Lang.

Deontological Comp.

– Performative Structure – Normative Structure

Agent1 Agent2 External agents msg

Norm. State
Norm. State'
Obligation

Rule EI DC DF

SLIDE 10

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 10/27

Electronic Institution: other

– EI = < DF, DC >

Governors

– Mediators among EI

and External Agents

– filter out forbidden

messages

Staff Agents

– EI delegates certain

tasks to these domain dependent agents.

– Their code is

implemented or trusted by the EI designer.

External agents Govern. Govern. DF / DC Staff Staff EI Agent1 Agent2

SLIDE 11

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 11/27

Situated El: objective

– Objective:

to enhance an

existing MAS (World) with EI regulation capabilities.

– Situate an EI in

a world (SEI)

interaction is

performed out

f EI
sense and act

SEI World Agent1 Agent2 situated sense act agent interaction

SLIDE 13

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 13/27

Situated El: objective

– Objective:

to enhance an

existing MAS (World) with EI regulation capabilities.

– Situate an EI in

a world (SEI)

Agent

interaction is performed out

f EI

SEI World Traffic scenario

EI = Traffic

Regulation Authority

Traffic code

violations

Sanctions
Collisions

situated sense act

SLIDE 14

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 14/27

Situated El: Modeller

SEI=<DF',DC,B>

Dialogical Frame.

– Modellers

model external

agents

– Properties

Agent (PA)
Institutional (PI)
Environment (PE)

SEI World Modeller (Car) DF' / DC

PI PApos PE

Staff (Signals)

SLIDE 15

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 15/27

Rule inform(police, car, noStop)

Situated El: Norms

SEI World Modeller (Car) DF' / DC Staff (Signals) Staff (Police)

PI PApoints PApos

Norm. State'
Obl(police, car, fine)

SEI=<DF',DC,B>

Deontological C.
Norms:

– Given SEI cannot

avoid certain interactions => add Rules to specify conseq.

SLIDE 16

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 16/27

Situated El: Bridge

Bridge

SEI=<DF',DC,B>

Bridge

– Implementation issue – Layer of abstraction

between world and SEI internal components.

– Distributed

SEI World Modeller (Car) DF' / DC Staff (Signals) Staff (Police)

PI PApoints PApos

SLIDE 17

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 17/27

Autonomic EI: objective

– Objective:

to autonomously adapt

an EI to achieve some defined goals.

– Autonomic EI (AEI)

Adaptation based on

goal fulfilment by extending an EI

Components:

– Institutional Goals (G) – observed Properties (P) – reconfiguration m. (TF)

AEI EI Interaction Agent1 Agent2 External agents

SLIDE 19

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 19/27

Autonomic EI: Goals

– AEI=<DF,DC,G,TF>

Institutional Goals (G)

– Goal Specification:

range of desired

values for observed Properties (P)

to measure properties

– Objective Function:

overall goal

satisfaction (weighted aggregation function)

to decide if adaptation

is required (using a threshold)

AEI EI G P I Agent1 Agent2 External agents

SLIDE 20

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 20/27

Autonomic EI: TF

AEI EI G

TF

P PS',NS' PS,NS I Agent1 Agent2 External agents

– AEI=<DF,DC,G,TF>

Transition Funcs. (TF)

– NS adaptation

e.g. increase fine amount

– PS adaptation

e.g. increase #policemen

: P×G×NS  NS' : P×G×PS  PS '

SLIDE 21

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 21/27

Related work: MoiseInst [Gateau06]

Govern. Dialogical F.:

roles

Deontological C.:

PS / Scenes
NS

EI External agents Wrapper Mng Structural Spec.

roles (w/ comp.

& author.) Contextual Spec. Normative Spec. Functional Spec. MoiseInst Agent StructMng. ContextM. NormMng. FunctMng. External agents Agent Instit.Mng. SceneMng. TransMng. EI and MoiseInst are very similar Situatedness and Autonomic exts. to MoiseInst seem feasible DomainD. NormMng.

SLIDE 23

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 23/27

Related work: other

– Situatedness:

Participant perception:

– Situated MAS [Weyns&al.]:

restrict agent perceptions depending on their virtual location.

– EASI, Cartago...

Organisation perception:

– SEI = EI awares of world – Simulator Bridge (interaction

is still inside EI)

– Adaptation:

Participant:

– Emergence of Norms

[Sen&Airau]

– Autonomous norm

compliance [Lopez&al.]

Organisation:

– AEI: org. adapts by itself,

having its own institutional goals.

SLIDE 24

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 24/27

Related work: COIN

– [Cardoso et al.08]: A Context-based Institutional

Normative Environment

To provide an enhanced Normative System to EI.

– [Lacroix et al.08]: The Use of Norms’ Violations to

Model Agents’ Behavioral Variety

Traffic Scenario; to generate agent population from rule

parameters.

– [Silva et al.08]: A Hybrid Reputation Model Based

n the Use of Organizations
To add a Reputation Service to EI.

SLIDE 25

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 25/27

Conclusions and future work

– SAEI: coordination

support system

Supervision of Social

conv.

Social conv.

Adaptation

– Future:

to provide assistance
iObjects: obj. with

delegated control capacities.

SAEI World Modeller Car DF' / DC Staff Signals Staff Police Bridge

PI PApoints PApos PE

Staff AdaptMng G P PS',NS' PS,NS

SLIDE 27

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 27/27

Thanks for your attention Questions?

SLIDE 28

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 28/27

SLIDE 29

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 29/27

Extra slides

SLIDE 30

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 30/27

Electronic Institution: DC

– EI = < DF, DC >

Dialogical Framework

– Ontology – Information Models – Social Structure – Communication Lang.

Deontological Comp.

– PS: Performative Structure – NS: Normative Structure

Agent1 Agent2 External agents EI DF DC NS PS

SLIDE 31

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 31/27

Electronic Institution: Scene

– EI = < DF, DC >

Dialogical Framework

– Ontology – Information Models – Social Structure – Communication Lang.

Deontological Comp.

– Performative Structure – Normative Structure

Agent1 Agent2 External agents EI DF DC

SLIDE 32

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 32/27

Electronic Institution: PS

– EI = < DF, DC >

Dialogical Framework

– Ontology – Information Models – Social Structure – Communication Lang.

Deontological Comp.

– Performative Structure – Normative Structure

Agent1 Agent2 External agents msg1 Governor Governor msg2 Scene A Scene B Scene C EI

SLIDE 33

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 33/27

Electronic Institution: NS

– EI = < DF, DC >

Dialogical Framework

– Ontology – Information Models – Social Structure – Communication Lang.

Deontological Comp.

– Performative Structure – Normative Structure

Agent1 Agent2 External agents Governor Governor msg1 NormStateA NormStateB

Prohibition

Rule EI

SLIDE 34

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 34/27

Individ. to collect.: Situated

EI Environment (for agents) situated sense act Agent1 Agent2 Existing MAS (≡ Environment for EI) Environment (for agents) SEI situated Agent1 Agent2 situated sense act situated agent interaction agent interaction situated

rg./agent

interaction External agents

EI: agents are situated in an

environment provided by EI. (also traditional Situated MAS)

SEI: institution as a whole

is situated in an existing MAS.

SLIDE 35

Formalising Situatedness and Adaptation in Electronic Institutions (COIN - AAMAS'08) 12/05/08 35/27

Individ. to collect.: Adapt.

Agent adaptation External agents Agent External agents adaptation AEI Agent External agents Social Conv. adaptation

– Individual level:

Agent reconfigures

itself

– individual goals and

properties – Org. level:

organisation (AEI)

reconfigures itself

(may be by staff agents)

– org. goals and

properties – Org. level by indiv.:

Agent reconfigure

the organisation

(eg. change constitution)

– Goals and

properties can be

indiv. and org.

EI Social Conv. EI Social Conv. Gorg Gind G

Formalising Situatedness and Adaptation in Electronic Institutions

Jordi Campos1, Maite López-Sánchez1,

Contents

Introduction: Context

– Organisations in MAS

Introduction: Motivation

– Motivation:

to enhance an existing MAS.

dynamic set of participant agents.

– Problem:

– Our approach:

level two agent properties:

Contents

Electronic Institution

– EI = < DF, DC >

Electronic Institution: DF

– EI = < DF, DC >

Electronic Institution: PS

– EI = < DF, DC >

Electronic Institution: NS

– EI = < DF, DC >

Electronic Institution: other

– EI = < DF, DC >

and External Agents

messages

tasks to these domain dependent agents.

implemented or trusted by the EI designer.

Contents

Situated El: objective

– Objective:

existing MAS (World) with EI regulation capabilities.

– Situate an EI in

a world (SEI)

performed out

Situated El: objective

– Objective:

existing MAS (World) with EI regulation capabilities.

– Situate an EI in

a world (SEI)

interaction is performed out

Situated El: Modeller

SEI=<DF',DC,B>

agents

Situated El: Norms

SEI=<DF',DC,B>

avoid certain interactions => add Rules to specify conseq.

Situated El: Bridge

SEI=<DF',DC,B>

between world and SEI internal components.

Contents

Autonomic EI: objective

– Objective:

an EI to achieve some defined goals.

– Autonomic EI (AEI)

goal fulfilment by extending an EI

Autonomic EI: Goals

– AEI=<DF,DC,G,TF>

values for observed Properties (P)

satisfaction (weighted aggregation function)

is required (using a threshold)

Autonomic EI: TF

TF

– AEI=<DF,DC,G,TF>

e.g. increase fine amount

e.g. increase #policemen

: P×G×NS  NS' : P×G×PS  PS '

Contents

Related work: MoiseInst [Gateau06]

Related work: other

– Situatedness:

restrict agent perceptions depending on their virtual location.

is still inside EI)

– Adaptation:

[Sen&Airau]

compliance [Lopez&al.]

having its own institutional goals.

Related work: COIN

– [Cardoso et al.08]: A Context-based Institutional

Normative Environment

– [Lacroix et al.08]: The Use of Norms’ Violations to