or and decision aidingbouyssou/orp3_bouyssou.pdforp3 – september 2005 – 4 some unanswered...
TRANSCRIPT
ORP3 – September 2005 – 1
OR and Decision Aiding
Denis BouyssouCNRS – LAMSADE
Some (personal) thoughts on our field and its future
ORP3 – September 2005 – 3
Operational ResearchYoung discipline
ORQ, Vol. 1, n° 1, March 1950Many scientific and practical achievements… butA discipline not without problems…
Without pressThat has not worked on its historyThat has been vigorously attacked
The future of Operational Research is past, JORS, 1979, R. L.Ackoff: “In my opinion, OR is dead even though it has yet to beburied”
That started as being applied and interdisciplinary andevolved differentlyGame Theory, Decision under uncertainty
ORP3 – September 2005 – 4
Some Unanswered Questions
What is OR?What does it mean to be interdisciplinary?What does it mean to be an applied science?Does this call for a specific epistemology?
How to teach OR and train OR analysts?What should be the standard curriculum in OR?
Should we remain isolated or try to ally with some sisterdisciplines (Computer Science, Statistics, Management, etc.)?
ROADEF: around 200 members
Should OR societies try to organize the profession?“Labeled” OR consultants?
ORP3 – September 2005 – 7
OR main modelBuild and solve models of the type
Solving:Finding an optimal solutionFinding all optimal solutionsFinding solutions “close” to theoptimal solutionsFinding robust solutions
How are such models built?What is their relation to realdecision processes?
ORP3 – September 2005 – 8
Example
Workforce Scheduling at Aéroport de ParisHigh cost of laborStrong unionsMany legal and technical regulationsDemand highly variable throughout timeHighly competitive industry
One category of persons“bagagistes”persons taking care of luggage from planes to customers
ORP3 – September 2005 – 9
0
20
40
60
80
100
1205 6 8 10 11 13 15 16 18 20 21 23
Demand Curve (normal weekday)
ORP3 – September 2005 – 11
Solve the problem
Minimiser x
s c
a x C t T
x h H
hh
H
ht hh
Ht
h
=
=
∑
∑ ≥ =
≥ =
1
11 2
0 1 2
. .
, , ,
, , ,
…
…
ORP3 – September 2005 – 12
����������
��������
����������
����������
��������
������������
������������
������������
������������
������������
����������
���������
����������
���������
���������
����������
����������
����������
����������������������������
������������
������������
������������
������������
���������������
����������
����������
����������
����������
�������������������
����������
����������
����������
����������
���������������
����������
����������
����������
����������
����������
�������������������
����������
����������
����������
��������
����������
����������
����������
�������������������
����������
����������
����������
����������
����������
��������
����������
��������
����������
����������
���������
���������
���������
���������
����������
����������
��������������������������������������
��������������������������������������������������������������
������������
������������
��������
����������
�����������������������
�����������
�����������
����������
����������
����������
���������������������������
������������
������������
����������������
����������
����������
����������
����������
����������
����������
����������
����������
����������
����������
����������������
����������
����������
����������
����������
�����������������������������������������
����������
����������
����������
����������
����������
����������
�����������������������������������
������������
������������
�����������
�����������
�����������
����������
����������
����������
����������
����������
�����������
�����������
����������
����������
����������
����������
����������
����������
����������
�����
020406080
100120140
5 6 8 10 11 13 15 16 18 20 21 23
An optimal solution
ORP3 – September 2005 – 13
“Data”
Definition of candidates shiftsImplicit Constraints
Legal / Union agreements / Habits / ConvenienceHard / Soft constraints
Dealing with perturbationsIllness / Late / Strikes / Overload
In practice?Who decides? When?Affectation of shifts to persons (auctions, seniority, etc.)?
Days / Weeks / MonthsOne model or several models?Combination?
ORP3 – September 2005 – 14
“Data”
Demand curveHow is it built?
Implicit Rulestype of plane × destination × time of the day
Implicit StandardsQuality?What is an acceptable waiting time?
Coping with uncertaintyLate planesChanges of planes
ORP3 – September 2005 – 15
What is the “problem”Why is it raised now?
MalfunctioningOpportunity
What objective(s)?Increased productivity?
How to share the gains?Negotiations?
Possible ShiftsTimetable of fights
ImplementationWho handle the problem?How is it handled?
Who will use the model?How often?To do what?
ORP3 – September 2005 – 16
Use of optimization modelOptimal solution is not necessarily central
Many unforeseen constraintsMany constraints difficult to modelHigh uncertainty / Many implicit aspectsMultiple heuristic adjustments of the solution
Model as an exploration toolcurrent practices / frontiers of system / objectives
Indirect use of modelPossible, frequent and legitimateModel as a tool to understand and/or dialogue
Necessary contextualization of solutions outside themodel
ORP3 – September 2005 – 17
In summary
OR analysts have a very naive strategy ofinterventionOR has no specific doctrine of implementation
ORP3 – September 2005 – 18
Strategy of the OR analyst
Faced with the complexity of real decisionprocesses and organizations, the answer of the ORanalysts is essentially pragmatic et non-formalized
Workforce scheduling (what to formalize and what toleave outside the model?)
Two “heroic” hypothesesIt is possible to have a positive role in organizationsusing a decision model that is essentially individual(optimization)Use “common sense” and “know how” to deal withimplementation issues
ORP3 – September 2005 – 19
Doctrine (lack of)
OR is an applied science…But OR analysts have not elaborated a doctrine ofintervention let alone an autonomous discourse onorganizations
How to justify the naive strategy?How did we react to the transformation of organizations?What have we learnt on organizations?How to train OR analysts
Supply chain, IS, Marketing, Strategic Management
ORP3 – September 2005 – 21
OR is not an isolated caseAim of OR
put reason and efficiency in the conduct of humanaffairs
Scientific Management (1910)OR (1950)Strategic Management (1960)Expert Systems (1970)MRP (1980)Supply chain Management (1990)Knowledge Management (2000)
ORP3 – September 2005 – 22
Models as “rational myths”
Unavoidable conflict between the logic of the modeland the logic of the organizationThe model is a porte d’entrée in the organization anda lever for change
OR models are good for doing so because they are rationalmyths
Myth: “Symbolic fable that is both simple and striking”Rational: “Performance” + high internal consistency
Models used unfreeze and recompose organizationalstructures
Learning / Contextualization / Implementation
Hatchuel & Molet, 1986, EJOR
ORP3 – September 2005 – 23
What does this view bring to OR?
Historical perspectiveSuccesses and failures of other rational myths
Doctrine: modeling and interventionModeling process
LearningConstructivism
ImplementationNot always requiredBeware of “compulsive implementation”
ORP3 – September 2005 – 24
Anatomy of a “rational myth”Rationalization project: 3 main components
Formal structureWhat is the underlying logic?
Management perspectiveWhat aspects of organizations are concerned?
Simplified view of organizationsWhat is the implicit view of organizations?
ORFormal structure
Pure decision theoryManagement perspective
Rationalize decisionsOrganization
Organizations as collections of decision makers
ORP3 – September 2005 – 25
Management perspectives:Evolution of rational myths
Work : TaylorDecisions : ORCosts : AuditMarkets : Strategic ManagementData : EIS / Data MiningFlows : Supply chainQuality : TQMProducts : ECRKnowledge : AI / KM?? : ??
ORP3 – September 2005 – 26
OR in context
OR is part of a long history of “rationalizationprojects” in organizations
Competition between “rationalization projects”We may and should learn from other fieldsThey should be able to learn from our field
What to do?
ORP3 – September 2005 – 28
Short termPay attention to competing “rational myths”
IS, Supply chain, TQM, AI, ECR, Data Mining, etc.INFORMS vs. EUROERP and OR
Work on history of OR
Pay attention to the evolution of organizationsDoctrine on organizations1st step: Who asks for OR?
Legal evolutions (deregulation / privatization)Technical evolutions (transports / NICT)
Classic techniques/models do not define ORThe nature of the rationalization project should govern the choice oftechniques
ORP3 – September 2005 – 29
Short termBe proactive: “We know how to handle that”
Prevention of industrial risksEvaluation of public policiesElectoral reformsSecuritySocial Security reformsGlobalizationClimate changeetc.
Or do we want to leave that to others? (economists,political scientists, sociologists, etc.)
Are they really better equipped than we are?
ORP3 – September 2005 – 30
OR’s competitive advantages
OR survived fashion effects50 years of experience
OR remains close to production processesScientific developments
Global Optimization, Heuristics, Complexity, Polyhedralanalysis, etc.
Prestigious journalsManagement Science, Operations Research
ORP3 – September 2005 – 31
Long Term
Formal structure: Pure decision theoryManagement perspective: Rationalizing decisionsOrganizations: Collections of Decision Makers
Theoretical challengeWhat is the present state of decision theory?
Practical challengeIs decision the right concept?
ORP3 – September 2005 – 32
Pure decision theory
The very idea of a “rational behavior” undergoes amajor crisis in decision theoryGoing from individual to collective model is aformidable challenge
ORP3 – September 2005 – 33
Independence axiom
z1
1/2
1/2
x
y
is preferred to
If
then (for p > 0)
is preferred to
z
k
p
1–p
1/2
1/2 x
y
k
p
1–p
ORP3 – September 2005 – 34
Independence axiom: Mom’s choice
Peter1
1/2
1/2
Peter
Mary
Preferred to
1/2
1/2
Peter
Mary
Preferred to
1/2
1/2 Mary
1/2
1/2
Peter
Mary
BUT
ORP3 – September 2005 – 35
(1 ; 1)
A
B
(0 ; 3)
(2 ; 2)
Games with perfect andcomplete information
(1 ; 1)
[A ; B]
ORP3 – September 2005 – 36
There are other possible and interestingmanagement perspectivesThe bulk of management literature is not aboutdecisionsIt is not at all obvious that “decisions” are centralany more in organizations
Rationalizing decisions?
ORP3 – September 2005 – 37
How CEO spend their time?15 % transports5 % visits5 % individual meetings30 % internal collective meetings10 % external collective meetings10 % meals15 % telephone5 % mail2 % writing2 % reading1 % thinking alone (6 min /day)0 % computer usage
20 %
55 %
25 %
ORP3 – September 2005 – 38
Rationalizing collective action
Deciding
StructuringPlanningCoordinatingControllingInventing incentivesContractingManaging risks
ORP3 – September 2005 – 39
Possible (necessary?) widening ofOR’s rational myth
What decision should I take?How to
Be organizedCoordinate with othersPlan while remaining reactiveControl resultsManage risksRemain informed
in order to be efficient?
ORP3 – September 2005 – 40
Possible New Key Words
ProductionConception / Quality /Environment
PlanningControl / Incentives /Contracts
SchedulingEquity / Motivation /Robustness
MarketsCompetition / Pricing /Strategy
TransportsSustainable development /Spatial equity
Rationalizing collectiveaction
Politics