29
Models and Frameworks David Levinson

Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Embed Size (px)

Citation preview

Page 1: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Models and Frameworks

David Levinson

Page 2: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

ALL MODELS ARE WRONG,

SOME MODELS ARE LESS WRONG THAN OTHERS.

• Some Models are Useful

Page 3: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

• Model - Mathematical Representation of a System

• Framework - Qualitative Organizing Principle for Analyzing System

Page 4: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Frameworks

Firm CustomersSuppliers

Competitors

Complements Regulation,Customs,Culture

© Levinson 2001

Page 5: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Diamond of Advantage

Page 6: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Example: Porter’s Diamond

• Michael Porter proposes four key determinants of competitiveness, which he calls the "Diamond of Advantage," based on cases from around the world:

1. factor conditions, such as a specialized labor pool, specialized infrastructure and sometimes selective disadvantages that drive innovation;

2. home demand, or local customers who push companies to innovate, especially if their tastes or needs anticipate global demand;

3. related and supporting industries, specifically internationally competitive local supplier industries, creating a high quality, supportive business infrastructure, and spurring innovation and spin-off industries; and

4. industry strategy/rivalry, involving both intense local rivalry among area industries that is more motivating than foreign competition and as well as a local "culture" which influences individual industries' attitudes toward innovation and competition.

Page 7: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Why Model

• gain insight into complex situations by understanding simpler situations resembling them

• optimization• system operation• learn from model building process• modeling as negotiation tool

Page 8: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Modeling Shapes Your Worldview, And Vice Versa• What is a worldview?

– Your outlook on life, and the world– Your internal model of how the world works

(I.e. what do you expect, what is a surprise)– The result of “Where you stand depends on

where you sit”

• Point of View– Who are the results for?– Subjective advocacy vs. objective analysis

Page 9: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Types of Models

• Network analysis• Linear Programming• Nonlinear Programming• Simulation• Deterministic queuing• Probabilistic queuing• Regression• Neural Nets• Genetic Algorithm• Cost/ Benefit Analysis• Life-cycle costing

• System Dynamics• Control Theory• Difference Equations• Differential Equations• Probabilistic Risk Assessment• Supply/Demand Equilibrium• Game Theory• Statistical Decision Theory• Markov Models• Cellular Automata• Etc.

Page 10: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Modeling Decisions• Hierarchy of Models• Scale/Detail• Time Frame• Spatial Extent• Boundaries (Boundary

Effects)• Macroscopic vs. Microscopic

(Zones, Flows vs. Individuals, Vehicles)

• Static vs. Dynamic• Stochastic vs. Deterministic• Linear vs. Nonlinear• Continuous vs. Discrete • Numerical Simulation vs.

Closed Form Solution

• Behavioral vs. Aggregate Model

• Physical vs. Mathematical Models

• Real Time vs. Offline• Short Term vs. Long Term

(Partial vs. General Equilibrium)

• Proactive vs. Reactive (Predictive vs. Responsive)

• Centralized vs. Decentralized (Optimization (Global) vs. Agent, Local Optimization)

• Equilibrium vs. Disequilibrium

Page 11: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Solution Techniques

• Understanding the System• Approximations and Speed in

Optimization (Local Optima) vs. Certainty (Brute Force, Global Optima)

Page 12: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Tradeoffs (Time and resource constraints)

• Money, • Data, • Computation, • Labor, • Ease of Use, • Convincing (e.g. Graphic Displays), • Extendable, • Evidence of Model Benefits, • Measuring Model Success

Page 13: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

PRT: Skyweb Express

Page 14: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

PROBLEM

• The Metropolitan Council of Governments (the region's main transportation planning agency) is examining whether the Twin Cities should build a new Personal Rapid Transit system in downtown Minneapolis, and they have asked you to recommend how it should be analyzed

• 1. What kind of model(s) should be used. Why?• 2. What data should be collected. • Form groups of 3 and take 10 minutes and think about

what kinds of models you want to run and what data you want to collect, what questions you would ask, and how it should be collected. Each group should have a note-taker, but all members of the group should be able to present findings to the class.

• DISCUSS

Page 15: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Some History

• A. World War II– i. deployment of radar in

a coordinated way– ii. spread to other fields

such as fighter tactics, mission planning and weapons evaluation.

– iii. use of mathematical techniques in such problems came to be known as operations research, other statistical and econometric techniques are being applied

• B. Post World War II– i. techniques spread to

universities. Mathematical development and application to a broad variety of problems.

– ii. development of systems analysis. Same analytical framework but to more complex problems for existing mathematical techniques were not adequate.

Page 16: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Some Definitions

• "A coordinated set of procedures which addresses the fundamental issues of design and management: that of specifying how men, money and materials should be combined to achieve a higher purpose" De Neufville

• "... primarily a methodology, a philosophical approach to solving problems for and for planning innovative advances" Baker

• "Professionals who endeavor to analyze systematically the choices available to public and private agencies in making changes in the transportation system and services in a particular region" Manheim

• "Systems analysis is not easy to write about: brief, one sentence definitions frequently are trivial" Thomas

Page 17: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

The Rational Planning Process

Define Problem, System (Constraints,Inputs, Outputs, Functions, Va lues,Evaluation Criteria)

Generate Solutions

Analyze Solutions

Evaluate Alternatives

Select Alternative

Implement

Define Goals and Objectives

Page 18: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Rational Planning ReduxSearch for Alternatives

Abstraction intoModel/Framework

Does Model Solution Exist?

Reconsider MOE

Insight/Knowledge

Selection

Evaluation

Prediction of Performance

1

3

2

Page 19: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Step A. Define the System

• i. objectives - measures the effectiveness or performance

• ii. environment - things which affect the system but are not affected by it

• iii. resources - factor inputs to do the work• iv. components - set of activities or tasks of the

system• v. management - sets goals, allocates resources

and exercises control over components• vi. model of how variables in 1-5 relate to each

other

Page 20: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Step B Generate and assess alternatives available to

management • 1. algorithms-systematic search over available

alternatives– analytical– exact numerical– heuristic numerical

• 2. generate alternatives selectively, evaluate subjectively– fatal flaw analysis– simple rating schemes– Delphi exercises

• 3. generate alternatives judgmentally, evaluate scientifically using system model

• 4. need for "stopping rule"

Page 21: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Step C Choose alternatives

• The analyst is generally not the decision maker. The actual influence of the results of the analysis in actual decisions will depend on:– determinacy of evaluation– confidence in the results on the part of

the decision maker– consistency of rating among alternatives

Page 22: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Step D Implementation

• Just Do It.

Page 23: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Step E Evaluation

• 1. definition: output from a later step in systems analysis used as input to a later step.

• 2. Examples.– analysis leads to revisions in systems

definition– implementation experience leads to a

revision of output system definition or values that underlay that definition.

Page 24: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Is the Rational Planning process Rational?

• [Discuss in pairs for ten minutes, be prepared to discuss in class]

Page 25: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Some Issues

• Limited Computational or Solution Generating Capacity

• Incomplete Information• Cost of Analysis• Conflicting Goals/Evaluation

Criteria• Reliance on Experts (What about

the People?)

Page 26: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Alternative Planning Decision Making

Paradigms. ¿Are They Irrational?

• Satisficing• Incrementalist• Organizational Process• Poltical Bargaining• Decomposition/Hierarchical

Strategy/Tactics/Operations

Page 27: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful

Summary

• 1. Applied systems analysis is the use of rigorous methods to assist in determining optimal plans, designs and solutions to large scale problems through the application of analytical methods.

• 2. Applied systems analysis focuses upon the use of methods, concepts and relationships between problems and the range of techniques available. Any problem can have multiple solutions. The optimal solution with depend upon technical feasibility (engineering) and costs and valuation (economics).

• 3. Applied systems analysis is an attempt to move away from the engineering practice of design detail and to integrate feasible engineering solutions with desirable economic solutions. The systems designer faces the same problem as the economist, "efficient resource allocation" for a given objective function.

Page 28: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful
Page 29: Models and Frameworks David Levinson. ALL MODELS ARE WRONG, SOME MODELS ARE LESS WRONG THAN OTHERS. Some Models are Useful