applying the ufo ontology to integrate organizational modeling viewpoints renata s.s. guizzardi...
TRANSCRIPT
Applying the UFO Ontology to Integrate
Organizational Modeling Viewpoints
Renata S.S. [email protected]
i* Internal WorkshopBarcelona, Spain
July, 2010
Outline Introduction:
Providing Knowledge Management through the agent-oriented paradigm
KM Systems Development Methodology: Combining Tropos + AORML
Ontology-based Approach How UFO was applied to combine Tropos and AORML
Ongoing Work in this direction KM Processes Development Methodology
Combining Tropos + ARIS Supporting Goal Elicitation through NFR Catalogues Goal Taxonomy to support alignment of goals and
processes
Knowledge Management
KM can be defined as a systematic process for acquiring, organizing and communicating
knowledge to all members of an organization, enabling them to be more effective and
productive in their work.
Conveying to people the right piece of information
at the right place, in the right time.
“There is no time for filling in the system with new knowledge.”
“Oh, it’s too much effort to fill in the system, and then I can never find something useful in it when I need it.”
“What if someone does something wrong with the knowledge I give away?”
“Why should I share my knowledge if knowledge is power?”
Knowledge Management Pitfalls
Lack of trust
Effort vs. Knowledge availability
Detachment from daily working practices
Lack of motivation
Theoretical Framework
knowledge spiral
communities of practice
distributed knowledge management
situated learning
genetic epistemology
social-historical constructivism
constructionism
dialogue and context in learning
knowledge management theories
constructivism
physical meaningful
artifactsperturbation context
autonomynon-hierarchical
knowledge sharingsocial interaction
Applying Agents as a Modeling Paradigm
Concepts which are able to capture the human dimension which are closer to people (communication to the
user is made easier)
belief
goalintention
social ability desires
learning ability
Combining Different Agent-oriented Modeling Languages
Methodology Requirements Developing a good understanding of the
organizational setting before jumping into the solution space.
Designing the system with an enough amount of detail to enable coding
Developing Effective KM Solutions
overall organizational goals
stakeholders goals
negotiating and reconciling
these goals
Methodology
Requirements Model + Architectural Design + Detailed Design
Tropos AORML
ARKnowD =
ARKnowD: Agent-oriented Recipe for KM Systems Development
Tropos’ Language Actor Resource Goal Softgoal Plan Dependency Decomposition Contribution Means-end
Characteristics of Tropos Potential
gives particular attention to requirements engineering, this makes it a natural candidate for organizational modeling
is based in goal modeling: represent organization’s and stakeholders’ goals
provides an abstract view of the organization (actors, goals, dependencies…), allowing us to leave details for later development cycles
Limitations does not provide tools to model agent’s interaction and
behavior with an appropriate amount of detail due to large use, constructs are extremely overloaded
(there is no consensus regarding their use)
Agent-Object-Relationship Modeling Language (AORML) Agent Object Action (Communicative Action) Interaction Event Commitment/Claim Association Specialization Composition Do/Perceive (the action)
Characteristics of AORML Potential
offers the means to model agent’s information, interaction and internal behavior in detail
naturally captures reactive behavior by using rules models both agents and objects provides deontic modeling constructs such as
commitments and claims, which form the basis for the establishment of such norms and contracts.
Limitations lacks constructs specific for requirements analysis limited case tools support so far
How to consistently combine Tropos and AORML? As they are applied in distinct development
activities, we can map Tropos to AORML using an MDD-inspired approach.
Ontology-based approach to evaluate and map the notations: We extended the UFO ontology to represent social
and intentional concepts present in the Agent domain.
We used the ontology to evaluate Tropos’ notation and AORML, correcting a few limitations of each.
We applied the ontology to understand which Tropos concepts mapped into the concepts in AORML.
Ontology applied to evaluate modeling languages (1/5)
Use the ontology as a reference model to which the language’s metamodel must be isomorphic to.
Language’s Metamodel
Reference Ontology
Ontology applied to evaluate modeling languages (2/5)
One concept in the language represents more thanone concept from the reference ontology
Reference Ontology
Language’s Metamodel
construct overload
Ontology applied to evaluate modeling languages (3/5)
Two concepts in the language represent the same concept from the reference ontology
Reference Ontology
Language’s Metamodel
construct redundancy
Ontology applied to evaluate modeling languages (4/5)
A concept in the language has no counterpart in thereference ontology.
Reference Ontology
Language’s Metamodel
construct excess???
Ontology applied to evaluate modeling languages (5/5)
A concept in the reference ontology has no counterpart in the language’s metamodel.
Reference Ontology
Language’s Metamodel
incompleteness???
Ontology applied to map two modeling languages
Map the concept C1 of language A into the concept C2 of language B in a way that C1 and C2 reference the same concept in the reference ontology.
Reference Ontology
Language A’s Metamodel Language
B’s Metamodel
map
UFO-C: Social and Mental Moments
UFO-C: Dependency vs. Delegation
Fixing Incompleteness in Tropos
Fixing Construct Overload in AORML
Transformation Rules
• MDD metamodel transformation: from CIM to PIM
ARKnowD
dependeedepender
goal dependency
delegateedelegator
goal delegation
dependeedepender
resource dependency
aquisiteeacquisitor
resource acquisition
Early Requirements
Late Requirements
human and institutional agents
plan
artificial agents
Architectural Design
Detailed Design
Transformation
agent
object
<<communication>>
relationships
Mike's AM: Artifact Model
Mike: Peer
createTaxonomy
title=”Health Insurance Taxonomy”
1
inclConceptTax
taxID= 12parent=nullconcept=”Health Insurance”
2
inclConceptTax
taxID= 12parent=”Health Insurance”concept="Client”
3
4
5
newTaxonomytaxID=12title="Health Insurance”
newConcepttaxID=12parent=nullconcept=”Health Insurance”
6
newConcepttaxID=12parent=”Health Insurance”concept="Client”
Interaction modeling
searchAnswer
Peer Assistant Artifact Manager
KeyQuestion,VectConcR1
<<belief>>Artifact Model
provideDocument Docs
provideExplanation
Question, Answer
noAvailableArtifact
behavior modeling
Tropos AORML
Ongoing Work Evolving UFO-C
ARKnowD’s Case Tool Completing previous work on implementing the
transformation from Tropos to AORML on an existing tool (TAOM4E).
Extending the methodology to coding (MDD: PIM to PSM).
Organizational Patterns (Semi)-automatically recognizing the Constructivist
KM Principles in the Tropos models.
But...
KM support does not always require a supporting system.
Business Process Modeling
Business Process Modeling focuses on a detailed understanding of the chain of
activities that deliver the organization’s products and services.
Main benefits Allowing traceability between goals and
business process models How goals are operationalized into BP. How BP impact the achievement of goals.
Providing Modularity both to Goal and BP models.
Diagnosing needs for reengineering.
Developing process-oriented information systems which are aligned with organization’s goals.
Combining Goals and BPM
Goal modeling + Business Process Modeling
Tropos ARIS-EPC
Organizational Model =
Example - BPMCarries out & Supports Carries out & Supports Carries out & Supports Carries out & Supports Carries out & Supports Carries out & Supports Carries out & Supports Carries out & Supports Carries out & Supports Carries out & Supports Other
Org
aniz
ati..
..
Patient Physician
Laboratorialexams
previouslyachieved
Verify thepreviously
achievement oflaboratorial exams
Laboratorialexams
not previouslyachieved
Ask forlaboratorial
exams
Performrequired
laboratorialexams
Laboratorialexams
performed
Process
A Fragment of a Business Process Model in ARIS: Diagnosing a Patient
BPM Approaches Neglect Goals
Minimize patient'sphysical sufferingand symptoms
Heal thepatient
Diagnosepatient's
health state
Prescribepatient`streatment
Show empathyto patient
Providemedical care
to patient
Overriding Goals
Subordinate Goals
Function ARIS goal model
Limitations of goal models of BPM approaches Do not allow an in depth goal analysis
Unclear semantics for decomposition. It does not model alternatives. It does not allow one to reason about how a
goal directly impacts other goals.
Weak connection to processes Relation about goals and processes is not
clear.
Lack of methodological guidance to elicit and model goals.
Atualmente: Tropos+ARIS
dependeedepender
goal dependency
delegateedelegator
goal delegation
dependeedepender
resource dependency
aquisiteeacquisitor
resource acquisition
Requisitos Iniciais
Requisitos Finais
human and institutional agents
plan
VAC EPC
Transformation
FAD
?
?
?
?
Tropos - objetivos ARIS - processos
artificial agents
Case Study A case study in a real organization was
conducted with the purpose of supporting the investigation regarding the relations between goals and processes.
Three phases: Elicitation phase: goal models and BPMs were
captured; Harmonization phase: a goal taxonomy was
created to help in the alignment of goals and BPs; Alignment phase: UFO is applied to clarify the
semantics of the elements of both models, enabling the alingment.
Elicitation Phase Preliminarly, standard methods were applied:
interviews and observation of work.
process oriented goals
Non Functional Requirements (NFR) Catalogues were applied, helping to elicit
allowed a more strategic point of view
NFR Catalogue (Chung et al., 2000)
Adjusting NFRs BP Requirements
NFRs have been originally proposed for system requirements elicitation. We should adjust them for eliciting BP requirements.
Approach: translating NFRs to the medical goal domain, relating the existing NFR types to selecting goals in our models.
One big distinction: originally, they lead to Tropos softgoals in our case, they may lead both to Tropos goals and
softgoals.
A Model without NFR
A Model with Catalogue
A few examples Accessibility - Access patient’s data records; Confidentiality - Maintain healthcare information private; Completeness - Obtain complete information about patient’s
treatment; Accuracy - Obtain accurate information about patient’s treatment; Traceability - Obtain traceability for information in patient’s
treatment (refined into Obtain traceability in investigation of patient’s condition, Obtain traceability in relation to treatment administered to patient and Obtain traceability in relation to physicians who prescribed patient’s treatment);
Integrability - Integrate service with other hospital departments, Integrate service with municipal and state health services and Integrate service with specialists in areas related to rheumatology;
Trust and confidence to the provider (assurance) - Trust physician
Empathy – provide patient with caring and personalized attention
Harmonization Phase
Taxonomy to guide how goals connect to processes (or portions of processes)
Total of 15 different goal types, classified according to 6 dimensions.
Examples: Dimension: Level of abstraction
Fundamental goal (provide medical care to patient) Process goal (diagnose patient health state) Activity goal (prescribe patient’s treatment)
Dimension: Temporal Aspect AS-IS (approve the treatment proposed by the resident) Change goal (standardize diagnosis cue sheets) TO-BE (coordinate patient care with other healthcare providers)
Acknowledgements
This research is funded by the Brazilian ResearchFunding Agencies FAPES (grant number 45444080/09) and
CNPq (grants number 481906/2009-6)