Information Economics in Design

1
Information Economicsin Design

• Chris Paredis
• The Systems Realization Laboratory
• PLM Center of Excellence
• G.W. Woodruff School of Mechanical Engineering
• Georgia Institute of Technology

www.srl.gatech.edu www.marc.gatech.edu/plm
2
What is Information Economics?

• Economics
• Study of the production, distribution and
  consumption of goods and services, and the
  management of these processes
• Study of how people choose to allocate scarce
  resources to satisfy competing uses or wants
• A study of choice
• Design
• Transformation of information from requirements
  to product description
• Information Economics in Design
• Which information should be created to support
  design decisions?
• What is the value of information? What is the
  cost of information?
• How can one generate more valuable information at
  a lower cost?

3
Foundations of Information Economics

• Some history
• Daniel Bernoulli (1738) Expected utility
• Knight (1921) Risk and uncertainty in economics
• von Neumann Morgenstern (1944) Utility theory
  
• Marschak (1950s) Economics of organization and
  information
• Renewed interest in the context of Information
  Systems (1990s)
• Value of information the difference in the
  expected value of a decision made with or without
  considering the information

4
Overview of Presentation

• Context
• What is Information Economics?
• Information and Knowledge in Product Development
• Examples of Information Economics in the SRL
• Related to Information
• How should one represent information and
  uncertainty?
• How should one use uncertain information to make
  decisions?
• How should one compute with uncertain
  information?
• Which information should one gather?
• Which models should one use?
• Related to Knowledge
• How should one represent knowledge, models?
• How should one manage knowledge, models?
• How should one design the design process?

5
Product Development A Decision-Based Perspective
Decisions
GenericDecisionProcess
6
Information-Driven Product Development
R
7
A Process Perspective
Product Perspective
Process Perspective
Process Order in which Relationships are Applied
8
Product Lifecycle Management Framework
9
Research Issues

• We need to develop a deeper understanding ofthe
  structure of the PLM information graph
• Which concepts relationships? ? Ontologies
• How to represent information and knowledge? ?
  uncertainty, context,
• How to reconcile multiple ontologies? ?
  interoperability
• Reusable patterns? ? Knowledge Repositories
• We need methods for managing the PLM information
  graph(creating, sharing, modifying,)
• Which tools to create and modify info? ? maps to
  stakeholders
• In which order to build the graph? ? concurrent
  engineering
• How to coordinate among multiple stakeholders?
• How to maintain consistency?
• How to propagate changes?
• How to maintain, retrieve and apply reusable
  knowledge templates?

10
Research Issues

• We need an IT infrastructure for distributed
  computation and collaboration support
• How to integrate multiple simulation, analysis,
  and optimization tools in a distributed fashion?
• Interoperability, security, load balancing,
• How to provide geographically distributed
  decision makers with relevant information in
  real-time?

11
Overview of Presentation

• Context
• What is Information Economics?
• Information and Knowledge in Product Development
• Examples of Information Economics in the SRL
• Related to Information
• How should one represent information and
  uncertainty?
• How should one use uncertain information to make
  decisions?
• How should one compute with uncertain
  information?
• Which information should one gather?
• Which models should one use?
• Related to Knowledge
• How should one represent knowledge, models?
• How should one manage knowledge, models?
• How should one design the design process?

12
How should one represent information and
uncertainty?(Jason Aughenbaugh, Scott Duncan)

• Aleatory uncertainty
• Inherently random irreducible
• Best represented as probability distribution
• Examples
• Manufacturing variability
• Epistemic uncertainty
• Due to a lack of knowledge
• Best represented as interval
• Examples
• Error due to model approximation
• Future design decisions
• Choose the representation that results in best
  design decisions

13
Probability Bounds Analysis P-boxes(introduced
by Ferson and Ginzberg, 1996)

• Combines probability distributions and intervals
• P-box Upper and Lower bound on all plausible
  CDF's
• Generalization of both intervals and probability
  distributions

To judge the value of the representation, one
needs to relate it to decisions
14
How should one make decision with P-boxes?(Jason
Aughenbaugh, Steve Rekuc)

• Expected Utility Interval !!
• Maps to set-based design
• Eliminate only the dominated designs
• Acknowledging ignorance results in better
  decisions !
• Characterize difference in performance
• Many sources of uncertainty are 'shared'
• Taking dependence into account reduces
  uncertainty in the difference in performance

Expected Utility
UB
LB
DV
Conservative Solution
Diff in Expected Utility
UB
LB
DV
Make better decisions with the same information
15
Which information to gather or models to
use?(Jay Ling)

• If epistemic uncertainty is too large to make a
  decision
• Gather more information
• Perform additional simulations (model
  information source)
• Perform the action that yields the most bang for
  your buck
• Satisficing solution
• When making a better decision costs more than it
  is worth
• Optimal in terms of Information Economics

Expected Utility
UB
LB
DV
Expected Utility
DV
Gather additional information most efficiently
16
Overview of Presentation

• Context
• What is Information Economics?
• Information and Knowledge in Product Development
• Examples of Information Economics in the SRL
• Related to Information
• How should one represent information and
  uncertainty?
• How should one use uncertain information to make
  decisions?
• How should one compute with uncertain
  information?
• Which information should one gather?
• Which models should one use?
• Related to Knowledge
• How should one represent knowledge, models?
• How should one manage knowledge, models?
• How should one design the design process?

17
How should one representing uncertain
knowledge?(Rich Malak)
Stress
sUB
ApplicabilityDomain
Strain
0
Epistemic Uncertainty
Goal Enable sharing and reuse of models
Amortize costs
18
Reusable and Composable Models(Manas Bajaj, Greg
Mocko, Nsikan Udoyen)

• Common associations between geometry and
  analyses/ simulations
• Common patterns between CAD description and
  simulation models

ReusablePatterns?
Enable reuse of models Amortize costs
19
Port-Based Abstraction Knowledge Templates

• Port
• Location of intended interaction
• Exchange of energy, material, signal
• Abstraction becomes container for associated
  models

Store knowledge in modular, reusable templates
Amortize costs
20
Reusable, Declarative Decision Templates(Marco
Fernandez, Jitesh Panchal)
Pressure Vessel
Spring
21
Summary

• Information Economics
• A framework for making decisions about design
• Applies to many of the problems we are working on
  in SRL
• Can serve as a guide for new research directions
• Which information costs dominate? How can we
  reduce the costs?
• How can we improve value?
• Questions? Comments?