The Integration of Para-consistent Conceptual Models Influenced by Uncertainty: A Belief-theoretic Approach

1
The Integration of Para-consistent Conceptual
Models Influenced by UncertaintyA
Belief-theoretic Approach

• Ebrahim Bagheri
• E.Bagheri_at_unb.ca

Supervisor Dr. Ali A. Ghorbani
2
Agenda

• Introduction
• Background and Related Work
• Challenges
• Motivations
• Proposed Model
• Contributions to Knowledge
• Evaluation Methods
• Thesis Timeline

3
Introduction

• Software development is a problem solving process
• The dimensions of the problem are
• Context
• Environment
• Situation
• Users
• etc.
• Solving a problem requires suitable tools!

4
Introduction (cntd.)

• Conceptual models are those tools
• Conceptual modeling is also known as problem
  analysis, or conceptualization
• Useful for the formalization of experts
  understanding of a problem
• Also useful for the systematic classification of
  knowledge and procedures

5
Why Conceptualize?

• Make real-world concepts tangible
• Support communication and collaboration
• Detect missing information, errors, or even
  misinterpretations
• Specify software behavior
• Provide an orientation on software performance

6
The Process
Early
Late
Problem Domain
Conceptualization
Product
7
Famous Models

• Software Engineering
• Classical Telos, DFD, SADT, ERD,
• Advanced UML, EM, KAOS, i,
• Knowledge Engineering
• Model, CML, KARL, CommonKADS

8
Viewpoint-Based Models

• Different areas and amount of knowledge helps
  better analyze a problem
• Employing more information sources for getting a
  better insight
• High complexity of a software system requires
  collaborative design effort
• Sources of information (participants) are known
  as viewpoints

9
(No Transcript)
10
Related Work

• Viewpoint-Oriented Requirement Definition (VORD
  Preview)
• Employ any notation
• Formal definition of Viewpoint
• viewpoint name
• viewpoint focus (boundary and scope)
• viewpoint concern (e.g. organizational goals,
  business objectives, etc.)
• viewpoint information sources
• viewpoint requirement definitions
• viewpoint activity history

11
Related Work (cntd.)

• Nuseibeh et al. define a viewpoint as
• Representation style
• Domain of interest (area of concern)
• Requirement specification
• Work plan (requirement engineering strategy)
• A work record
• Inflexible due to strict declaration of
• Representation style
• Work plan

12
Related Work (cntd.)

• Controlled Requirement Expression (CORE)
• divide the problem domain into disjoint areas of
  concern
• Consistent and complete model when sub-models are
  merged into a single model
• How can we clearly split the problem domain?

13
Challenges to VP-based Models

• Defining a unique representation style
• Use of common terminology (vocabulary)
• Identifying specification overlaps
• Detecting model discrepancies
• Merging different models
• Evaluating the final product

14
Challenges (cntd.)

• Humans make conception errors due to
• Risk aversion
• Short term memory
• Perceptual problems
• Epistemic uncertainty (aka partial ignorance)
• Not all information sources are equally reliable
• Different problem statement strategies

15
Current Solutions

• Shared ontology
• Common application vocabulary
• Thesauri
• Human expert inspection
• Conflict forms
• Evaluating conflict metrics
• Category theory
• Belnaps knowledge order

16
Current Solutions (cntd.)

• Formal Methods
• Convert viewpoint specifications into
• VWPI and use static analyzer
• First-order logic and use backward and forward
  chaining
• Goal regression in KAOS
• Incrementally elicited ranked structures
• Make use of epistemic states
• Apply preference orderings

17
Motivations

• Lack of attention to the following issues
• Need to capture uncertainty
• Unify model integration for different CM schemas
• Address information source reliability
• Real-time model development, negotiation, and
  integration
• Evaluation of integration effectiveness

18
Uncertainty

• Dempster-Shafer theory of evidence
• A tool for numerically quantifying and reasoning
  under uncertainty
• An extension to Probability theory where power
  set elements receive probability (belief) mass
• A rich area of research with tools for
• Belief function combination (e.g. Dempsters rule
  of combination)
• Belief propagation
• Belief entropy measures (e.g. generalized entropy)

19
Proposed Model
Conceptual Modeling Layer
Underlying Construct Layer
Belief Layer
20
Schematic View
Update
Update
Shared Understanding (Consensus)
Viewpoint
Viewpoint
Update
Viewpoint
21
Sample Process
22
Belief Space(Subjective Logic)
Uncertainty
1
Highly Uncertain
Very Uncertain
0
0
Slightly Certain
Highly Certain
1
1
Belief
Firm Disagreement
Slight Disagreement
Slight Agreement
Firm Agreement
Either Way
Disbelief
0
23
Sample Process (ctnd.)
24
Contributions to Knowledge

• Employing linguistic terms
• Addressing uncertainty
• Creating an underlying modeling construct
• Consensus building and negotiation
• Pre-consensus belief recommendation

25
Contributions to Knowledge (cntd.)

• Merging conceptual models
• Evaluating merging effectiveness
• Information source reliability assessment
• Tool support for the proposed model

26
Evaluation Methods

• Formally prove the correctness of the
• Merge operators
• Belief recommendation algorithms
• Analyze the behavior of the operators under
  extreme conditions
• Conduct several case studies using
• A number of Computer Science graduate students
• A tool that will support the proposed solution
• The preliminary tool is being developed in
  Eclipse (EMF)

27

• Thank you ?

28
Goal regression

• The idea of goal regression is linked to planning
  via plan modification...i.e. in order to achieve
  P and Q, construct a plan F that achieves P, and
  then modify F so that it achieves Q while still
  achieving P. The idea is to protect P so that the
  choice of where to place the steps for achieving
  Q is determined relative to the plan for P.