Week 7-A working model – and then?

A Core Course on Modeling
Week 7-A working model – and then?
     Contents     
• The Need for Interpretation
• Approach
• Criteria for Modeling
•Genericity
•Scalability
•Specialization
•Audience
•Convincingness
•Distinctiveness
•Surprise
•Impact
• Examples
• Criteria for modeling and purposes
1
A Core Course on Modeling
Week 7-A working model – and then?
     The Need for Interpretation     
2
define
context  initial problem
conceptualize
initial problem  conceptual model
formalize
conceptual model  formal model
execute
conclude
formal model  result
result  resolve initial problem?
A Core Course on Modeling
Week 7-A working model – and then?
     The Need for Interpretation     
3
A Core Course on Modeling
Week 7-A working model – and then?
     Approach     
4
Quallity assessment needs comparison; comparison needs criteria
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for Modeling     
5
Try formulate orthogonal criteria, applying to all possible situations
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for Modeling     
Property 1:
Criterion regards the begin (definition stage) or the end
(conclusion stage)
Example ‘definition’: what is the problem scale? (scalability)
Example ‘conclusion’: how much depends on it? (impact)
6
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for Modeling     
Property 2:
Criterion regards ‘inside’ (model, modeled system) or
‘outside’ (stakeholders + context)
Example ‘inside’: how convincing is the model outcome?
(convincingness)
Example ‘outside’: how much need stakeholders to know?
(specialization)
7
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for Modeling     
8
Property 3:
Criterion regards ‘qualitative’ or ‘quantitative’ aspects of the
model / modeled system
Example ‘quantitative’: how large is the intended audience?
(audience)
Example ‘qualitative’: how different can modeled systems
be? (genericity)
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for Modeling     
Combinations
define/
conclude
inside /
outside
9
scalability
qualitative /
quantitative
criterion
define
inside
qualitative
genericity
define
inside
quantitative
scalability
define genericity
audience
specialization
distinctiveness
define
define
outside
outside
qualitative
quantitative
conclude
inside
qualitative
conclude
inside
quantitative
conclude
outside
qualitative
conclude
outside
quantitative
specialization
audience
convincingness
distinctiveness
surprise
impact
conclude
convincingness
inside
impact
surprise
qualitative
outside
quantitative
A Core Course on Modeling
Week 7-A working model – and then?
     Genericity (define,inside,qualitative)     
Genericity
10
A Core Course on Modeling
Week 7-A working model – and then?
     Genericity (define,inside,qualitative)     
Genericity
what variety of types of modeled systems can be captured?
11
A Core Course on Modeling
Week 7-A working model – and then?
     Genericity (define,inside,qualitative)     
12
Genericity
Formula for volume 6-face.
cube, side h:
V1 = h3
rectangular block, height h:
V2 = h*Areatop
truncated pyramid:
V3 = h*(Areatop + (Areatop*Areabottom) + Areabottom)/3
A Core Course on Modeling
Week 7-A working model – and then?
     Scalability (define,inside,quantitative)     
Scalability
13
A Core Course on Modeling
Week 7-A working model – and then?
     Scalability (define,inside,quantitative)     
Scalability
how does performance relate to problem size n
O(1)
O(log n)
O(n)
O(n log n)
O(np)
O(2n), O(n!), …
14
A Core Course on Modeling
Week 7-A working model – and then?
     Scalability (define,inside,quantitative)     
Scalability
how does performance relate to problem size n
O(1)
O(log n)
O(n)
O(n log n)
O(np)
O(2n), O(n!), …
n
15
A Core Course on Modeling
Week 7-A working model – and then?
     Scalability (define,inside,quantitative)     
Scalability
how does performance relate to problem size n
O(1)
n
O(log n)
n
O(n)
O(n log n)
O(np)
O(2n), O(n!), …
16
A Core Course on Modeling
Week 7-A working model – and then?
     Scalability (define,inside,quantitative)     
Scalability
how does performance relate to problem size n
O(1)
n
O(log n)
n
O(n)
n
O(n log n)
O(np)
O(2n), O(n!), …
17
A Core Course on Modeling
Week 7-A working model – and then?
     Scalability (define,inside,quantitative)     
Scalability
how does performance relate to problem size n
O(1)
n
O(log n)
n
O(n)
n
O(n log n)
n
O(np)
O(2n), O(n!), …
18
A Core Course on Modeling
Week 7-A working model – and then?
     Scalability (define,inside,quantitative)     
Scalability
how does performance relate to problem size n
O(1)
n
O(log n)
n
O(n)
n
O(n log n)
n
O(np)
n
O(2n), O(n!), …
19
A Core Course on Modeling
Week 7-A working model – and then?
     Scalability (define,inside,quantitative)     
Scalability
how does performance relate to problem size n
O(1)
n
O(log n)
n
O(n)
n
O(n log n)
n
O(np)
n
O(2n), O(n!), …
n
20
A Core Course on Modeling
Week 7-A working model – and then?
     Specialization
(define,outside,qualitative)
Specialization

21
A Core Course on Modeling
Week 7-A working model – and then?
     Specialization
(define,outside,qualitative)

22
Specialization
how much specialized knowledge should problem owner have
how much control should problem owner have over
• validity of assumptions,
• regime of application,
• use of results?
A Core Course on Modeling
Week 7-A working model – and then?
     Specialization
(define,outside,qualitative)

23
Specialization
how much specialized knowledge should problem owner have
1. Invest in presentation
• graph style
• labels, captions, scales
• uncertainty
• combine plots
A Core Course on Modeling
Week 7-A working model – and then?
     Specialization
(define,outside,qualitative)

24
Specialization
how much specialized knowledge should problem owner have
2. when in doubt,
be conservative
A Core Course on Modeling
Week 7-A working model – and then?
     Specialization
(define,outside,qualitative)

25
Specialization
how much specialized knowledge should problem owner have
3. be warned for biased users
• detailed account of assumptions
• insist: report is indivisible
• refrain from easy-to-misapprehend results
• get a second opinion
A Core Course on Modeling
Week 7-A working model – and then?
     Specialization
(define,outside,qualitative)

26
Specialization
how much specialized knowledge should problem owner have
4. wicked problems:
• self-fulfilling
• self-denying
A Core Course on Modeling
Week 7-A working model – and then?
     Audience (define,outside,quantiative)    
Audience
27
A Core Course on Modeling
Week 7-A working model – and then?
     Audience (define,outside,quantiative)    
Audience
size of the intended audience
• Large size: low level of specialization
• Large size: no bi-directional communication
• Large size: consider interactive model (example:
‘stemwijzer’)
28
A Core Course on Modeling
Week 7-A working model – and then?
     Convincingness
(conclude,inside,qualitative)
Convincingness

29
A Core Course on Modeling
Week 7-A working model – and then?
     Convincingness
(conclude,inside,qualitative)

Convincingness
more convincing if fewer and/or less implausible assumptions
1. Assumptions logically deducible?
2. If not: first-principle ‘laws’?
3. If not: ‘formal’ model system?
4. If not: ‘empirical’ model system?
5. If not: argument for consistency of the assumptions?
30
A Core Course on Modeling
Week 7-A working model – and then?
     Convincingness
(conclude,inside,qualitative)

Convincingness
more convincing if fewer and/or less implausible assumptions
1. Assumptions logically deducible?
2. If not: first-principle ‘laws’?
3. If not: ‘formal’ model system?
4. If not: ‘empirical’ model system?
5. If not: argument for consistency of the assumptions?
31
A Core Course on Modeling
Week 7-A working model – and then?
     Convincingness
(conclude,inside,qualitative)

Convincingness
more convincing if fewer and/or less implausible assumptions
1. Assumptions logically deducible?
2. If not: first-principle ‘laws’?
3. If not: ‘formal’ model system?
4. If not: ‘empirical’ model system?
5. If not: argument for consistency of the assumptions?
32
A Core Course on Modeling
Week 7-A working model – and then?
     Distinctiveness (conclude,inside,quantitative)     
Distinctiveness
33
A Core Course on Modeling
Week 7-A working model – and then?
     Distinctiveness (conclude,inside,quantitative)     
34
Distinctiveness
more distinctive: distinguishes between more similar alternatives
Examples:
• prediction 1: how close can predicted times T1 and T2 be?
• explanation: how close can phenomena Q1 vs. Q2 be?
• optimization: how close can optima P1 and P2 be?
A Core Course on Modeling
Week 7-A working model – and then?
     Distinctiveness (conclude,inside,quantitative)     
35
Distinctiveness
more distinctive: distinguishes between more similar alternatives
two common types of errors:
false positive
(conclude X where there is no X)
false negative
(don’t conclude X where there is X)
A Core Course on Modeling
Week 7-A working model – and then?
     Surprise (conclude,outside,qualitative)     
Surprise
36
A Core Course on Modeling
Week 7-A working model – and then?
     Surprise (conclude,outside,qualitative)     
37
Surprise
the extent to which it may bring unforeseen new ideas
Open and closed spaces of outcomes.
Examples of closed outcomes:
• A model computing a probability of X produces 0…1
• A model verifying Y can only produce ´true´ or ´false´.
A Core Course on Modeling
Week 7-A working model – and then?
     Surprise (conclude,outside,qualitative)     
Surprise
the extent to which it may bring unforeseen new ideas
Open and closed spaces of outcomes.
Examples of open outcomes:
• ontologies
• evolutionary algorithms
• PCA, abstraction, …
38
A Core Course on Modeling
Week 7-A working model – and then?
     Impact (conclusion,outside,quantitative)     
Impact
39
A Core Course on Modeling
Week 7-A working model – and then?
     Impact (conclusion,outside,quantitative)     
40
Impact
the extent to which the model outcome can affect stakeholders
Two perspectives:
• prestige and profit: the more impact the better
• risk and responsability: the less impact the better
A Core Course on Modeling
Week 7-A working model – and then?
     Impact (conclusion,outside,quantitative)     
41
Impact
the extent to which the model outcome can affect stakeholders
capitalize intended impact:
r1=revenues ; no model outcome;
r2=revenues; outcome present;
c1=costs; no model outcome;
c2=costs; outcome present;
=((r2-r1)-(c2-c1))/(|r2-r1|+|c2-c1|)
A Core Course on Modeling
Week 7-A working model – and then?
     Impact (conclusion,outside,quantitative)     
42
Impact
the extent to which the model outcome can affect stakeholders
quantify adverted impact:
C=chance/time of an incident
V=value loss per incident
CV has dimension of money / time
add to c2 in the formula for .
A Core Course on Modeling
Week 7-A working model – and then?
     Impact (conclusion,outside,quantitative)     
43
Impact
the extent to which the model outcome can affect stakeholders
balance impact and reliability
A Core Course on Modeling
Week 7-A working model – and then?
     Example (an evening in the bar)     
Genericity:
extend the model to account for various types of drinks
http://dir.groups.yahoo.com/group/Beverage-Fiesta/
44
A Core Course on Modeling
Week 7-A working model – and then?
     Example (an evening in the bar)     
Scalability:
extend the model to
plan your behavior
for an entire year
(including building
up or destroying
friendships)
http://thetipsycrow.com/p-3009-calendar.html
45
A Core Course on Modeling
Week 7-A working model – and then?
     Example (an evening in the bar)     
Specialization:
tune the model to be
used by an AA-coach
(e.g., deal with ml
C2H5OH instead of nr.
glasses of beer)
http://thelastcolor.blogspot.nl/
46
A Core Course on Modeling
Week 7-A working model – and then?
     Example (an evening in the bar)     
Audience:
use the
model in a
website ‘plan
your evening
in the bar’
Stratumseind
1-
Eindhoven
http://skatepark-leidsenhage.hobbyfront.nl/bereikbaar_skatepark_leidsenhage.html
47
A Core Course on Modeling
Week 7-A working model – and then?
     Example (an evening in the bar)     
Convincingness:
base the amount of talktime per glass on
empirical data, e.i., do a
black box model first
http://www.cafepress.co.uk/+still_talking_drinking_glass,559284302
48
A Core Course on Modeling
Week 7-A working model – and then?
     Example (an evening in the bar)     
Distinctiveness:
calculate with number
of sips instead of
number of glasses
http://familyfitness.about.com/od/healthandsafety/tp/summer_safety.htm
49
A Core Course on Modeling
Week 7-A working model – and then?
     Example (an evening in the bar)     
50
Surprise:
incorporate digestion into the model, so that it can suggest you to
consume a plate of bitterballen when appropriate
http://www.franfoodservice.nl/snacks/index.htm
A Core Course on Modeling
Week 7-A working model – and then?
     Example (an evening in the bar)     
51
Impact:
have the model be used by traveling salesmen: assume the
amount of talk-time is proportional to the amount of acquisition,
and hence the amount of leads.
http://www.lemonlaws.com/buying_new_cars/
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
Which criteria go with which purposes?
Which purposes require which ciriteria?
52
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
prediction (1, 2):
• convincingness,
• distinctiveness,
• impact
53
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
compression:
• scalability,
• audience,
• distinctiveness
54
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
inspiration:
• surprise
55
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
unification:
• genericity,
• convincingness,
• surprise
56
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
abstraction:
• distinctiveness,
• convincingness,
• surprise
57
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
verification:
• scalability,
• convincingness,
• impact
58
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
exploration:
• genericity,
• surprise
59
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
decision:
• convincingness,
• distinctiveness,
• impact
60
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
optimization:
• genericity,
• scalability,
• convincingness,
• impact
61
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
specification:
• genericity,
• distinctiveness
62
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
realization:
• genericity,
• distinctiveness,
• impact
63
A Core Course on Modeling
Week 7-A working model – and then?
     Criteria for modeling and purposes     
steering & control:
• distinctiveness
64
A Core Course on Modeling
Week 7-A working model – and then?
     Summary     
65
•Leading question: to what extent has the initial problem been solved?
•Approach: criteria to assess the quality of the modeling process
•Taxonomy: definition or conclusion stage? inside or outside? Qualitative or quantitative?
•Resulting criteria:
•
Genericity: how many different modeled systems can we handle?
•
Scalability: how large can the size of the problem be?
•
Specialization: how much should the intended audience know?
•
Audience: how large can the intended audience be?
•
Convincingness: how plausible are the assumptions?
•
Distinctiveness: e.g., how accurate, how certain, how decisive is the model outcome?
•
Surprise: to what extent can the model outcome give new insight?
•
Impact: how big can the consequences of the model outcome be?
•Criteria for modeling quality are related to purposes.