conceptual modeling csc2507 intentions and agentsjm/2507s/notes04/goals.pdf2004 john mylopoulos...
TRANSCRIPT
2004 John Mylopoulos Intentions and Agents -- 1
Conceptual Modeling CSC2507
Intentions and AgentsIntentions and Agents
From Entities and RelationshipsFrom Entities and Relationshipsto Goals and Agentsto Goals and Agents
Intentions, Goals, Intentions, Goals, SoftgoalsSoftgoalsDesign RationaleDesign Rationale
Agents and Social SettingsAgents and Social SettingsSpeech Acts and Action WorkflowsSpeech Acts and Action Workflows
Actors and Social DependenciesActors and Social Dependencies
2004 John Mylopoulos Intentions and Agents -- 2
Conceptual Modeling CSC2507
Types of DomainsTypes of Domainsand Models Thereofand Models Thereof
�� Static modelsStatic models use concepts such as EntityEntity, AttributeAttribute,RelationshipRelationship, ResourceResource,...
�� Dynamic modelsDynamic models described in terms of ProcessProcess, ActivityActivity,ActionAction, PlanPlan, ProcedureProcedure, EventEvent,...or StateState, TransitionTransition,...
�� Intentional modelsIntentional models describe the worlddescribe the world of things agents(human or otherwise) believe in, want, prove, argueabout, e.g., IssueIssue, GoalGoal, SoftgoalSoftgoal, SupportsSupports,…
�� Social modelsSocial models describe social settings in terms of socialrelationships among agents, such as Authority,Commitment, Responsibil i ty, ActorActor, PositionPosition, RoleRole,Goal/Task/Resource DependencyGoal/Task/Resource Dependency,...
� ...Others...
2004 John Mylopoulos Intentions and Agents -- 3
Conceptual Modeling CSC2507
The State-of-the-ArtThe State-of-the-Art� Static models have been studied since the beginning of
conceptual modeling, e.g., Entity-Relationship model,...� Dynamic models have also been explored since the early
days of computer science, partly independently ofconceptual modeling, e.g., state machines, Petri nets,...
� Intentional models have seen less research; there has beenmuch work in AI planning, more recent work on issue-basedmodels of (software) design and goal-based RE,...
� Social models have been studied the least withinconceptual modeling, but are becoming important thanks torise of agent-oriented software systems.
...the action is with intentional and social models!...the action is with intentional and social models!
2004 John Mylopoulos Intentions and Agents -- 4
Conceptual Modeling CSC2507
Intentional ModelsIntentional Models
� Intentional models encompass the world of things agents(human or otherwise) believe in, want, prove, etc.
� Goals have been studied in AI since the ‘50s, mostly as partof a formal framework for doing planning.
� A goal is a desired state, often described in terms of apredicate,E.g., profits(year(2005)) ≥ $1B (strategic goal)Or sales(VW beetle,week(13/03/2000)) ≥ 5
(operational goal)
2004 John Mylopoulos Intentions and Agents -- 5
Conceptual Modeling CSC2507
AND/OR Goal GraphsAND/OR Goal Graphs
� Goals can be AND- or OR-decomposed to build AND/ORgraphs.
� A simple procedure exists for AND/OR graphs fordetermining whether a root node of an AND/OR goal graphis solved/fulfilled, given that some other nodes of the graphhave been found to be solved/fulfilled, or unsolvable/unfulfillable.
2004 John Mylopoulos Intentions and Agents -- 6
Conceptual Modeling CSC2507
Schedulemeeting
Collectconstraints
Collecttimetables
Collect other constraints
Share timetables
Systemcollects
Personcollects
By email
By all means
From all From
Initiatoronly
Manually Automatically
Interactively
Generateschedule
AND decomposition
OR decomposition
2004 John Mylopoulos Intentions and Agents -- 7
Conceptual Modeling CSC2507
A Critique of Planning-Type GoalsA Critique of Planning-Type Goals
� (Planning-type) Goals are formally defined; but some(real-world) goals may not be formally definable(remember primitive concepts in Classic?).
� (Planning-type) Goals are consistent; but goals may beconflicting, as in Requirements Engineering.
� Goals may contribute positively or negatively to eachother’s fulfillment, but such weak dependencies can’t berepresented at all in terms of AND/OR relationships.
� Some forms of goal analysis may be useful even if goalsare not fully formalized, see BPR and requirementsanalysis applications.
2004 John Mylopoulos Intentions and Agents -- 8
Conceptual Modeling CSC2507
SoftgoalsSoftgoals
� These are goals which, like primitive concepts in Classic,don’t have a formal definition. Consequently, softgoals don’thave a clearcut criterion as to whether they are fulfilled ornot (hence their name…)
� Softgoals are satisficed , rather than satisfied; in otherwords, softgoal fulfillment is relative and “good enough”,rather than absolute and optimal.
� Softgoals were introduced in [Mylopoulos92] and [Chung93]as a primitive concept for modeling non-functionalrequirementsE.g., User-friendly[Interface2],
Portable[Module4]
2004 John Mylopoulos Intentions and Agents -- 9
Conceptual Modeling CSC2507
Programmability
+
+
++Allow
Change of Colours Allow
Change ofState
Allow Change ofLanguage
ErrorAvoidance Information
Sharing Ease ofLearning
UserTailorability
Usability
User-DefinedWriting Tool
Modularity
UseComponents
UserFlexibility
Allow Change ofSettings
+
+
+
2004 John Mylopoulos Intentions and Agents -- 10
Conceptual Modeling CSC2507
Goal RelationshipsGoal Relationships� To arrive at a more qualitative framework for modeling
goals, we also need to extend the set of relationshipsbetween goals beyond AND- and OR-relationships:�+ -- one goal contributes positively towards the
fulfillment of another goal;� - -- one goal contributes negatively towards the
fulfillment of another goal;�++ (--) -- one goal subsumes/negates another, I.e., if
the first goal is fulfilled, the second is fulfilled/denied;� With these enhancements, we can build goal models
which could be useful for strategic business analysis orrequirements analysis (as opposed to planning).
2004 John Mylopoulos Intentions and Agents -- 11
Conceptual Modeling CSC2507
Building Goal Dependency GraphsBuilding Goal Dependency Graphs
� Start from one or more goals and/or softgoalsG1,G2,…,Gn which need to be fulfilled together.
� Analyze each, looking for ways to fulfill it through AND-or OR-decompositions, or through other refinementswhich contribute positively.
� Continue this process until there is enough positivesupport to fulfill all root nodes. At this point you have ndisconnected goal trees T(G1), T(G2),…,T(Gn).
� Identify positive and negative inter-tree influences, I.e.,positive or negative relationships between goals g, g’which belong to different goal trees.
� Repeat the analysis to see if root goals are fulfilled; if so,done, else continue the analysis.
2004 John Mylopoulos Intentions and Agents -- 12
Conceptual Modeling CSC2507
Flexibility
-
+-
-Use
Passwords
Availability
Do Backups
+
+ +
++
Security
+
Confidentiality
Usability
Integrity
Maintenability
Recoverability
Performance
Minimizeexternal
communication
Minimizeredundancy
Accuracy
Extratesting
2004 John Mylopoulos Intentions and Agents -- 13
Conceptual Modeling CSC2507
Softgoals Softgoals as Criteriaas Criteria
Schedulemeeting
By all means By
-
- +
++
+
-
-
Collecttimetables
By person
Bysystem
Have updatedtimetables
Collectthem
Chooseschedule
Manually
Automatically
MatchingeffortCollection
effort
Minimalconflicts
Degree ofparticipation
Quality ofscheduleMinimal
effort
2004 John Mylopoulos Intentions and Agents -- 14
Conceptual Modeling CSC2507
Issues,Positions and ArgumentsIssues,Positions and Arguments
� Understanding decisions often involves asking why?why? :• “Why is this variable declared in this block?”• “Why did we decide to sell our Oshawa plant?” etc.
� To answer such questions, we need to represent,somehow, the rationalerationale for particular decisions. Thisrationale may link particular decisions to operational goals,or operational goals to strategic ones, and strategic goalsto their origins (stakeholders etc.)
� This is an important issue for Software Engineering ingeneral, but is also important for strategic businessanalysis and other knowledge management applications.
2004 John Mylopoulos Intentions and Agents -- 15
Conceptual Modeling CSC2507
What do we Want to Represent?What do we Want to Represent?These are progressively more refined models for capturing
decision rationale:
�� Arguments spaceArguments space -- a decision is associated with allrelevant arguments, also their interrelations (e.g.,implies, supports, denies, qualifies)
�� Alternatives spaceAlternatives space -- alternatives and their argumentsare made explicit; arguments about a particularalternative are differentiated from other arguments.
�� Evaluation spaceEvaluation space -- evaluation measures are madeexplicit along with their arguments
�� Criteria spaceCriteria space -- now criteria used for evaluation aremade explicit as well, along with their arguments
�� Issues spaceIssues space -- issues are made explicit, along withtheir alternatives, evaluations, and criteria
2004 John Mylopoulos Intentions and Agents -- 16
Conceptual Modeling CSC2507
ToulminToulmin ’’s s Model of ArgumentationModel of Argumentation� Toulmin, a philosopher, proposed his model for describing
arguments in 1958 [Toulmin58]. His model is limited to theargument space.
Placing window Placing window commands in global commands in global
menu reducesmenu reducesscreen clutterscreen clutter
Global menuGlobal menucan becomecan becomevery largevery large
Commands needCommands neednot be duplicatednot be duplicatedfor each windowfor each window
No duplicationNo duplicationsaves spacesaves space
Common senseCommon sense
usuallyusually
DatumDatumClaimClaim
ModifierModifier
WarrantWarrant
BackingBacking
RebuttalRebuttal
soso
sincesince
unlessunless
onAccountOfonAccountOf
2004 John Mylopoulos Intentions and Agents -- 17
Conceptual Modeling CSC2507
IBIS andIBIS and gIBIS gIBIS� IBIS [Kunz70], the Issue-Based Information SystemIssue-Based Information System ,
is a first comprehensive attempt to capture decisionrationale in Computer Science.
� gIBIS is the graphical version of IBIS, based onhypertext ideas [Conklin87].
� Basic idea is to capture the issues that were raisedduring the decision making process, the arguments proand con particular positions and the resulting decision.
� Nodes of an IBIS network represent issuesissues , positionspositionsand argumentsarguments .
� Framework can be used to expose faulty assumptions,unexplored alternatives and weak or missing rationale.
2004 John Mylopoulos Intentions and Agents -- 18
Conceptual Modeling CSC2507
IBIS Representation of a DiscussionIBIS Representation of a Discussion
Issue1Issue1Issue1
Issue3Issue3Issue3
Issue2Issue2Issue2
Issue4Issue4Issue4
Position1Position1Position1
Position2Position2Position2
Argument1Argument1Argument1
Argument2Argument2Argument2
Argument3Argument3Argument3
Argument4Argument4Argument4
Argument5Argument5Argument5
generalizesrespondsTo
respondsTo
questions objectsTo
objectsTo
objectsTo
objectsTo
supports
isSuggestedBy
2004 John Mylopoulos Intentions and Agents -- 19
Conceptual Modeling CSC2507
IBIS Representation of a DiscussionIBIS Representation of a Discussion
Overcrowdedhospitals
OvercrowdedOvercrowdedhospitalshospitals
Peopletreated
in the US
PeoplePeopletreatedtreated
in the USin the US
Fundingfor socialprograms
FundingFundingfor socialfor socialprogramsprograms
HomelessHomelessHomeless
Morefunding
MoreMorefundingfunding
Cuttingtaxesbetter
CuttingCuttingtaxestaxesbetterbetter
Can’t run deficitsCanCan’’t run deficitst run deficits
No big government
No big No big governmentgovernment
What aboutpoor people?
What aboutWhat aboutpoor people?poor people?
generalizesrespondsTo
respondsTo
questions
objectsTo
objectsTo
objectsTo
objectsTo
supports
isSuggestedBy
No prooftax cuts
help
No proofNo prooftax cutstax cuts
helphelp
People want good medical insurance
People want good People want good medical insurancemedical insurance
2004 John Mylopoulos Intentions and Agents -- 20
Conceptual Modeling CSC2507
Potts andPotts and Bruns Bruns
� Proposal for capturing design rationale, IBIS-based.� Design rationale represented as a network of designdesign
artifactartifact and deliberationdeliberation concepts.� Design artifacts document (elements of) the design.� Deliberations are issuesissues , alternativesalternatives or justificationsjustifications ..�� PlanetextPlanetext is the graphical implementation of the system,
using hypertext links.� A Planetext network can actually be translated into Horn
clauses and fed into a Prolog engine which analyzes itlooking for inconsistencies.
[Potts88]
2004 John Mylopoulos Intentions and Agents -- 21
Conceptual Modeling CSC2507
A Potts andA Potts and Bruns Bruns Network Network
DDD
AAA
AAA
AAA
III
III
III
JJJ
JJJ
JJJ
DDD
DDD
DDD
AAA
AAA
AAADDD
III
AAA
JJJ
ArtifactArtifact
IssueIssue
AlternativeAlternative
JustificationJustification
DerivesDerives
TTTT
TT
TaskTask
2004 John Mylopoulos Intentions and Agents -- 22
Conceptual Modeling CSC2507
Potts and Potts and BrunsBruns : An Example: An Example
DDD
AAA
AAA
AAA
III
III
JJJ
JJJ
DDD
DDD
AAA
AAA
TT
TT
FormatterFormatter
ReadInputReadInput
InterpretInputInterpretInput
TT
ProduceOutputProduceOutput
ReadInputEncaReadInputEnca--psulationLevelpsulationLevel
HaveDocBufferHaveDocBuffer
DocLineDocLine
ReadInputInlineReadInputInline
InterpretInputEncaInterpretInputEnca--psulationLevelpsulationLevel
MakeDocLineAbstrMakeDocLineAbstr
IIIProduceOutputEncaProduceOutputEnca--psulationLevelpsulationLevel
DocDoc
2004 John Mylopoulos Intentions and Agents -- 23
Conceptual Modeling CSC2507
Questions, Options, Criteria (QOC)Questions, Options, Criteria (QOC)� Offers a framework for exploring design spaces, based
on the comparison of alternative options.� Extends Potts&Bruns framework with goalsgoals , criteriacriteria .� A QOC network is composed of:
��QuestionsQuestions -- pose key issues;
��OptionsOptions -- possible alternatives;
��CriteriaCriteria -- for selecting among options;
��ArgumentsArguments -- support/challenge a criterion.� DRL is an implementation of QOC. QOC does not
“capture design”; rather, it is supposed to be part ofto be part of thedesign process
[McLean91]
2004 John Mylopoulos Intentions and Agents -- 24
Conceptual Modeling CSC2507
CriteriaCriteria� A criterioncr iterion�Measures a property that the designer controlscontrols
indirectlyindirectly by choosing an option;�Must be unconditionalunconditional , i.e., the better the criterion,
the better the design;�Must be evaluativeevaluative in that it measures a property of
the artifact;� Criteria must be justified by referencing other criteria.
�� Bridg ing criteriaBridg ing criteria relate design decisions to generalcriteria, e.g., throughput vs performance.
�� Note: Note: Potts&Bruns issues are the result of a designdecision; QOC options are the result of a question(issue)
2004 John Mylopoulos Intentions and Agents -- 25
Conceptual Modeling CSC2507
QOC Diagram for ATM DesignQOC Diagram for ATM Design
Q: What range ofservices to offer?
Q: What range ofQ: What range ofservices to offer?services to offer?
Q: How to select cash amount?
Q: How to select Q: How to select cash amount?cash amount?
Q: Where to retrievecash and receipt from?
Q: Where to retrieveQ: Where to retrievecash and receipt from?cash and receipt from?
Q: How to initiate the transaction?
Q: How to initiate Q: How to initiate the transaction?the transaction?
O: Full rangeO: Full rangeO: Full range
O: Cash onlyO: Cash onlyO: Cash only
O: Same slotO: Same slotO: Same slot
O: Identify customerO: Identify customerO: Identify customer
O: Select cash amountO: Select cash amountO: Select cash amount
O: Same slot as receiptO: Same slot as receiptO: Same slot as receiptO: Different slotsO: Different slotsO: Different slots
O: DefaultsO: DefaultsO: DefaultsO: Typing and defaultsO: Typing and defaultsO: Typing and defaults
O: Type in amountO: Type in amountO: Type in amount
C: SpeedC: SpeedC: SpeedC: Variety of servicesC: Variety of servicesC: Variety of services
Q: How are services restricted?
Q: How are Q: How are services restricted?services restricted?
O: Fixed machinesO: Fixed machinesO: Fixed machines
O: Switchable machinO:O: Switchable Switchable machin machin
C: FastC: FastC: FastC: ObviousC: ObviousC: ObviousC: Variety of amountC: Variety of amountC: Variety of amount
C: FastC: FastC: FastC: Low costC: Low costC: Low cost
C: ObviousC: ObviousC: ObviousC: FastC: FastC: Fast
C: Low errorC: Low errorC: Low error
2004 John Mylopoulos Intentions and Agents -- 26
Conceptual Modeling CSC2507
Social Social OntologiesOntologies� These are collections of concepts for modeling social
settings, e.g., a family, a university, or a research group.� The units of a social setting are actors , such as agents ,
positions and roles . Actors may be atomic, such as Maria(...a person), InfoSleuth (…a program), or composite, suchas the recruiting committee, DCS, UofT.
� Social relationships are inherently intentional . This meansthat they constrain the actors they relate over manypossible worlds, rather than just the here-and-now world.
� For example, OnTopOf(projector,table) is extensional. ButMother(maria,george), Promise(attila,greg,a) are intentionalin that they constrain the behaviour of their actors in futurestates of the world. The same holds for relationships suchas owns, reportsTo,...
2004 John Mylopoulos Intentions and Agents -- 27
Conceptual Modeling CSC2507
Modeling Social SettingsModeling Social SettingsTwo complementary ways to model social settings:
�� Work Action ApproachWork Action Approach -- focuses on speech acts asmeans for changing the intentional state of actors; forexample, a request speech act, if successful, getsthe actor being requested to commit to fulfill therequest.
�� Social Dependency ApproachSocial Dependency Approach -- represent explicitlyactor goalsgoals and inter-actor dependenciesinter-actor dependencies and howthese constrain social activity. For example, amanager depends on her engineers to deliver designson time, while the engineers depend on the managerto give them a raise for their good work.
2004 John Mylopoulos Intentions and Agents -- 28
Conceptual Modeling CSC2507
The Action Workflow FrameworkThe Action Workflow Framework� Some communication actions are intended to change
the state of the social setting, such as�Asking someone for information;�Asking someone to have something done;�Having someone commit to do something etc.
� Such social actions are called speech actsspeech acts [Searle69].� Speech acts were originally proposed as a new
approach to Linguistics, claiming that some utterancescan best be understood as actions towards achievingsome goale.g., “It’s cold in here” means something like “someone,
turn up the thermostat, please” (…at least when aking says it in the presence of his court.)
2004 John Mylopoulos Intentions and Agents -- 29
Conceptual Modeling CSC2507
Speech Acts and ConversationsSpeech Acts and Conversations� [Winograd86] introduced speech act ideas to system
design by arguing that in designing a human-computersystem we need to focus on the interaction, rather thanthe human or the computer.
� Offered a substantial critique of AI and other areas ofcomputer science for focusing too much on one or theother side, rather than the interaction itself.
� Proposed to use speech acts to model such interactions.� This work was fundamental in establishing Computer-Computer-
Suppor ted Collaborative WorkSupported Collaborative Work (CSCW) as anautonomous research area within Computer Science, alsoproduced one of the first workflow products (thetheCoordi natorCoordinator ).
2004 John Mylopoulos Intentions and Agents -- 30
Conceptual Modeling CSC2507
The Basic Molecule of Social ActionThe Basic Molecule of Social Action
1 2 4 5
683
7 9
C: clientS: server
C:request S:promiseS:report
C: declare
C:complete
C: cancel
S:renege
S:withdrawS:rejectC:withdraw
S:declineC:withdraw
C: counter
S:counterC: accept
Think of this as the basic social action “molecule”, out ofwhich one can build social processes
2004 John Mylopoulos Intentions and Agents -- 31
Conceptual Modeling CSC2507
The Action Workflow FrameworkThe Action Workflow Framework
Could youCould you please do? please do?
Yes, IYes, I’’lllldo itdo it
ItIt’’s dones doneOK,OK,
thanksthanks
RequestorRequestor ActorActor
A simpler action molecule:
Asks forAsks foran actionan action
Agrees toAgrees todo itdo it
Fulfills work andFulfills work andreports it donereports it done
Accepts report Accepts report and declares and declares satisfactionsatisfaction
2004 John Mylopoulos Intentions and Agents -- 32
Conceptual Modeling CSC2507
Getting a New Credit AccountGetting a New Credit Account
Current accountCurrent accountcreditcredit
ClientClient BranchBranchClerkClerk
Current accountCurrent accountcreditcredit
ClientClient BranchBranchManagerManager
client is sentto the manager
acceptanceletter
communicationcredit is
operative
forms
consult bankDB for clientcredit record secondary workflows
for credit resolution
2004 John Mylopoulos Intentions and Agents -- 33
Conceptual Modeling CSC2507
Opening a New Credit AccountOpening a New Credit Account
Open Open current accountcurrent account
BranchBranchManagerManager
BranchBranchStaffStaff
CollectCollectclient dataclient data
BranchBranchStaffStaff
Enter dataEnter dataBanksBanks ’’ Risk RiskCentreCentre
BanksBanks ’’ Risk RiskCentreCentre
EDPEDPCentreCentre
2004 John Mylopoulos Intentions and Agents -- 34
Conceptual Modeling CSC2507
Complete Credit Account WorkflowComplete Credit Account Workflow
Manage creditManage creditlifecyclelifecycle
Audit creditAudit creditperformanceperformance
Produce DailyProduce DailyOverdraft RepsOverdraft Reps
Review creditReview creditaccountsaccounts
Clarification Clarification & explanation& explanation
docsdocs
Recover overdueRecover overduecreditscredits
AC -- Area CoordinatorAC -- Area CoordinatorBM -- Branch ManagerBM -- Branch ManagerIO -- Inspectors OfficeIO -- Inspectors Office
ACAC
ACAC
BMBM
BMBM
BMBM
BMBM
BMBM
IOIO
EDP Dept EDP Dept
Branch CreditBranch CreditOfficerOfficer
ClientClient
Legal DeptLegal Dept
2004 John Mylopoulos Intentions and Agents -- 35
Conceptual Modeling CSC2507
i*: Actors,i*: Actors,Dependencies and CommitmentsDependencies and Commitments
� Stands for “distributed intentionality” [Yu95].� Basic idea is that a social setting consists of actors who
have goals and depend on each other for their fulfillment.
�� ActorsActors have goalsgoals , need taskstasks to be carried out andresourcesresources to be made available; actors can be agentsagents(human or otherwise), rolesroles or positionspositions
�� DependenciesDependencies define intentional relationships amongactors, where one actor depends on another to satisfy agoal or a softgoalsoftgoal, execute a task; or furnish a resource.
� Dependencies can be cr it icalcr it ical , committedcommitted or openopen ,depending on the strength of commitment of the actorsinvolved and the amount of inconvenience caused by anunfulfilled dependency.
2004 John Mylopoulos Intentions and Agents -- 36
Conceptual Modeling CSC2507
The Strategic Dependency ModelThe Strategic Dependency Model� The key idea here is to focus on social dependenciesdependencies
among actors, rather than actor goals, actions etc.� Can be seen as complementary to the Work Action
framework: Work Action deals with the process throughwhich commitments and dependencies are generated,while the Strategic Dependency Model models existingcommitments through dependencies.
2004 John Mylopoulos Intentions and Agents -- 37
Conceptual Modeling CSC2507
The Strategic Dependency ModelThe Strategic Dependency Model
BodyShop
Owner
Appraiser
InsuranceCompany
Carrepaired
Pay repairs
Maximizeestimate
Continuebusiness
D D
D D
D D
DD
D D
D
D
D
D
D
D
D
D
Claimspayout
Premiumpayment
D D
Customerhappy
Repairscovered
D
D
Appraisedamages Minimize
repairs
Secureemployment
Fair repairappraisal
GoalGoal
TaskTask
ResourceResource
Soft GoalSoft Goal
2004 John Mylopoulos Intentions and Agents -- 38
Conceptual Modeling CSC2507
““ ...Let the Agent Handle it......Let the Agent Handle it... ””
BodyShop
Owner
InsuranceAgent
InsuranceCompany
Carrepaired
Pay repairs
Maximizeestimate
Continuebusiness
D D
D D
D D
DD
D D
D
D
D
D
D
D
Claimspayout
Premiumpayment
D D
Customerhappy
Repairscovered
D
D
D
D
Fair repairpayout
Goodrelations
Claimspayout
D
D
D
D
Claimsreimburse
D
D
Agencyrights
Small claimshandled
2004 John Mylopoulos Intentions and Agents -- 39
Conceptual Modeling CSC2507
““ ... Let the Body Shop Handle it...... Let the Body Shop Handle it... ””
Owner
Body Shop
InsuranceCompany
D
D D
D
D D
D
Premiumpayment
D
D
D
DD
D
ApprovedBody Shop
Owner notfraudulent
D
Repairsreimburse
Carrepaired
Customerhappy
Repairscovered
DD
Body shop notfraudulent
D
D
Small claimshandled
2004 John Mylopoulos Intentions and Agents -- 40
Conceptual Modeling CSC2507
The Need to Capture RationaleThe Need to Capture Rationale
How do actorsHow do actors ’’ goals lead to social dependencies? goals lead to social dependencies?�� Means-endsMeans-ends analysis can be used to relate goals to tasks
that can satisfy these goals: “Given goal (end) G, how canI decompose it (means) in order to find a way to fulfill it”.
� Means-ends analysis has been used in AI since GPS(General Problem Solver) proposed by Simon, Newelland Shaw in the ‘50s.
�� Task decompositionTask decomposition links relate tasks to other,component tasks
� Tasks can also be decomposed into goals.
2004 John Mylopoulos Intentions and Agents -- 41
Conceptual Modeling CSC2507
Means-Ends Analysis within theMeans-Ends Analysis within theStrategic Rationale ModelStrategic Rationale Model
Claim besettled
Verifypolicy
ClaimsHandlingClaimsClaims
HandlingHandling
Handleclaim
Settlementcost?
Prepareoffer
Whosefault?
Get accidentinfo
Determinefault
Police
WitnessDoctor
Appraiser
Determinecost to settle
Accidentinfo
Sufficienttreatment Injury
infoAppraisedamage
Minimalrepairs
DDD D
D
ActorActorboundaryboundary
2004 John Mylopoulos Intentions and Agents -- 42
Conceptual Modeling CSC2507
Functional AlternativesFunctional AlternativesClaim besettled
Verifypolicy
Handleclaim centrally
Prepareoffer
Agenthandles claim
Body Shophandles claim
alternativeone
alternativealternativeoneone alternative
twoalternativealternative
twotwo alternativethree
alternativealternativethreethree
2004 John Mylopoulos Intentions and Agents -- 43
Conceptual Modeling CSC2507
Non-Functional RationaleNon-Functional Rationalefor Choosing Among Alternativesfor Choosing Among Alternatives
Claim besettled
Agenthandles claim
Body Shophandles claim
alternativeone
alternativealternativeoneone
alternativetwo
alternativealternativetwotwo
alternativethree
alternativealternativethreethree
Handleclaim centrally
Profitable HappyCustomers
Low costs
Low litigation risk Fast claimsprocessing++
++++
++++
++--++
++-- ++++--
[Chung93]
2004 John Mylopoulos Intentions and Agents -- 44
Conceptual Modeling CSC2507
FormalizationFormalization� Actor dependencies are characterized using intentional
concepts, such as beliefsbeliefs , goalsgoals , commitmentscommitments , etc.� For example, here are the axioms for committed goalcommitted goal
dependency:CW(a, b, φ) ≡ CW(a, φ) ∧ B(a, CA(b, φ))CW(a, φ) ⇒ B(a, ∃ p, ∃ φ 0( ¬ φ ⇒ fail(a, p, φ 0 )))CA(a, φ) ⇒ B(a, ∃ p( result(p, φ) ∧ allDepOK(a, p)))
2004 John Mylopoulos Intentions and Agents -- 45
Conceptual Modeling CSC2507
TroposTropos : A Formal Extension of i*: A Formal Extension of i*� Although the primitive concepts of i* are defined
formally, i* is not an expressive language because itdoesn’t support the specification of constraints,invariants, pre/post-conditions for different elementsof an i* model.
� Tropos extends the expressiveness of i* , adoptingideas from KAOS.
� Most notably, the Tropos specification languageincludes a temporal logic inspired by KAOS.
� Actors, goals, actions, entities, relationships aredescribed both statically and dynamically.
� …stay tuned…
2004 John Mylopoulos Intentions and Agents -- 46
Conceptual Modeling CSC2507
�� [Begeman88[Begeman88] Begeman, M., Conklin, E., “The Right Tool for the Right Job”, Byte,October 1988.
� [Bunge77] Bunge,M.,Treatise on Basic Philosophy: Ontology I, The Furniture ofthe World, Reidel, 1977.
� [Chung93] Chung, L., Representing and Using Non-Functional Requirements: AProcess-Oriented Approach, PhD thesis, Department of Computer Science,University of Toronto, June 1993.
�� [Conklin87[Conklin87] Conklin, E., “Hypertext: An Introduction,” IEEE Computer, September1987.
�� [Conklin91[Conklin91] Conklin, E., and Burgess-Yakemovic, K., “A Process-OrientedApproach to Design Rationale”, Human-Computer Interaction 6(3-4), 1991.
�� [Gotel95][Gotel95] Gotel, O., Finkelstein, A., “Contribution Structures”, ProceedingsSecond IEEE International Symposium on Requirements Engineering, York,England, March 1995.
� [Gruber91] Gruber, T., “The Role of a Common Ontology in Achieving Sharable,Reusable Knowledge Bases,” Proceedings of the Second InternationalConference on Principles of Knowledge Representation and Reasoning,Cambridge, 601-602.
ReferencesReferences
2004 John Mylopoulos Intentions and Agents -- 47
Conceptual Modeling CSC2507
References References (cont(cont ’’d)d)
� [Guarino00] Guarino, N. and Welty, C., “Ontological Analysis of TaxonomicRelations,” in A. Länder and V. Storey (eds.), Proceedings of ER-2000: TheInternational Conference on Conceptual Modeling, Springer Verlag LNCS 1920.
� [Guarino01] Guarino, N. and Welty, C., “Conceptual Modeling and OntologicalAnalysis”, tutorial notes, 2001.
� [Kunz70] Kunz, W. and Rittel, H., “Issues as Elements of Information Systems”,Working Paper, Centre for Planning and Development Research, University ofCalifornia, Berkeley, 1970.
�� [Lee90[Lee90] Lee, J., "Sibyl: A Qualitative Decision Management System", in Winston, P.and Shellard,S. (eds.) AI at MIT: Expanding Frontiers, MIT Press, 1990.
�� [Lee91[Lee91] Lee, J. and Lai, K-Y, “What is Design Rationale?”, Human-ComputerInteraction 6(3-4), 1991.
� [Macfarlane95Macfarlane95] Macfarlane, I., and Reilly, I., “Requirements Traceability in anIntegrated Development Environment”, Proceedings Second IEEE InternationalSymposium on Requirements Engineering, York, England, March 1995.
�� [MacLean91[MacLean91] MacLean, A., Young, R., Bellotti, V. and Moran, T., “Questions,Options, Criteria: Elements of Design Space Analysis”, Human-ComputerInteraction 6(3-4), 1991.
2004 John Mylopoulos Intentions and Agents -- 48
Conceptual Modeling CSC2507
References References (cont(cont ’’d)d)
� [Mylopoulos92] Mylopoulos, J., Chung, L. and Nixon, B., "Representing and UsingNon-Functional Requirements: A Process-Oriented Approach", IEEE Transactionson Software Engineering, June 1992.
� [Potts88] Potts, C. and Bruns, G., “Recording the Reasons for Design Decisions”,Proceedings International Conference on Software Engineering, Singapore, 1988.
� [Searle69] Searle, J., Speech Acts: An Essay in the Philosophy of Language,Cambridge University Press, 1969.
� [Toulmin58] Toulmin, S., The Uses of Argument, Cambridge University Press, 1958.
� [Wand90] Wand, Y. and Weber, R., “An Ontological Model of an InformationSystem”, IEEE Transactions on Software Engineering 16, 1282-1292.
� [Winograd86] Winograd, T. and Flores, F., Understanding Computers andCognition: A New Foundation for Design, Addison-Wesley, 1986.
� [Yu95] Yu, E., Modeling Strategic Relationships for Process Reengineering, PhDthesis, Department of Computer Science, University of Toronto, January 1995.