Download ppt - Samer Hassan

Data-Driven Agent-Based Social Simulation
of Moral Values Evolution
Samer Hassan
Universidad Complutense de Madrid
University of Surrey
Contents

The Problem

ABM Mentat: Design

ABM Mentat: Results

AI: Fuzzy Logic

AI: Natural Language Processing

AI: Data Mining
Samer Hassan
SSASA 2008
2
Objective

Study the evolution of Spanish society in
the period 1980-2000

Data-Driven Agent-Based Modelling

Applying several Artificial Intelligence
techniques
Samer Hassan
SSASA 2008
3
The Problem

Aim: simulate the process of change in
moral values


in a period
in a society

Plenty of factors involved

Nowadays, centred in the inertia of
generational change:

Samer Hassan
To which extent the demographic dynamics
explain the mentality change?
SSASA 2008
4
The Problem

Input Data loaded: EVS-1980




Quantitative periodical info
Representative sample of Spain
Allows Validation
Intra-generational:


Samer Hassan
Agent characteristics remain constant
Macro aggregation evolves
SSASA 2008
5
Contents

The Problem

ABM Mentat: Design

ABM Mentat: Results

AI: Fuzzy Logic

AI: Natural Language Processing

AI: Data Mining
Samer Hassan
SSASA 2008
6
Design of Mentat

Agent:


 World:
EVS  Agent MS attributes
 3000 agents
 Grid 100x100
Life cycle patterns
 Demographic model

Demographic micro-evolution:
• Couples
• Reproduction
• Inheritance
 Network:
 Communication with
Moore Neighbourhood
 Friends network
 Family network
Samer Hassan
SSASA 2008
7
Friendship Network
Samer Hassan
SSASA 2008
8
Friendship Network
Samer Hassan
SSASA 2008
9
Friendship Network
Samer Hassan
SSASA 2008
10
Friendship Network
Samer Hassan
SSASA 2008
11
Friendship Network
Samer Hassan
SSASA 2008
12
Friendship Network
Samer Hassan
SSASA 2008
13
Friendship Network
Samer Hassan
SSASA 2008
14
Friendship Network
Samer Hassan
SSASA 2008
15
Methodological aspects

Data-driven ABM


Design with qualitative info


Microsimulation concepts
Life cycle, micro-processes
Introduction of empirical equations

Life expectancy, birth rate, different probabilities

Initialisation with survey data

Validation with different empirical data
Samer Hassan
SSASA 2008
16
Mentat in action
Samer Hassan
SSASA 2008
17
Contents

The Problem

ABM Mentat: Design

ABM Mentat: Results

AI: Fuzzy Logic

AI: Natural Language Processing

AI: Data Mining
Samer Hassan
SSASA 2008
18
Results
Samer Hassan
SSASA 2008
19
Results

It may arise new sociological knowledge:
Demographic Dynamics are a key factor for the prediction
of social trends in Spanish society
Samer Hassan
SSASA 2008
20
Contents

The Problem

ABM Mentat: Design

ABM Mentat: Results

AI: Fuzzy Logic

AI: Natural Language Processing

AI: Data Mining
Samer Hassan
SSASA 2008
21
Introduction of AI: Fuzzy Logic

Why Fuzzy Logic?


Social sciences are characterized by uncertain and vague
knowledge
Different concept than probability
Age
Young Adult
Old
10
1
0
0
20
0.8
0.8
0.1
30
0.5
1
0.2
40
0.2
1
0.4
50
0.1
1
0.6
Samer Hassan
SSASA 2008
22
Fuzzification

Attributes

Similarity

Friendship & its evolution

Couples
Samer Hassan
SSASA 2008
23
Contents

The Problem

ABM Mentat: Design

ABM Mentat: Results

AI: Fuzzy Logic

AI: Natural Language Processing

AI: Data Mining
Samer Hassan
SSASA 2008
24
Introduction of AI: NLP

Fuzzy logic helps for ABM qualitative input

NLP helps for ABM qualitative output

Experimenting with life-events generation:


Output in natural language: life-story of a representative
individual (Ex: hyper-inflation)
Applications:

NL format makes direct comparison with real stories possible

Information very simple for any individual to understand

Complementing explanations of quantitative research
Samer Hassan
SSASA 2008
25
Quantitative & Qualitative Output Generation
Macro
Trends
Quantitative
Statistics and
Graphs
European
Values
Survey
Micro
processes:
interactions
Simple
Filtering
Samer Hassan
Events
log
(XML)
Content
Determination:
Complex
filtering based
on rules
Analysis,
Filtering
and NLG
Discourse
Planning:
Ordering for a
coherent story
SSASA 2008
Life-Story of
Representative
Individual (ideal
type)
Surface
Realization:
Natural Language
Generation
26
An example: part of the XML output
<Log Id="i49">
<Description />
<Attribute Id="name" Value="rosa" />
<Attribute Id="last_name" Value="pérez" />
<Attribute Id="sex" Value="female" />
<Attribute Id="ideology" Value="left" />
<Attribute Id="education" Value="high" />
...
<Events>
<Event Id="e1" Time="1955" Action="birth" Param="" />
<Event Id="e2" Time="1960" Action="friend" Param="i344" />
<Event Id="e3" Time="1960" Action="friend" Param="i439" />
<Event Id="e4" Time="1961" Action="friend" Param="i151" />
<Event Id="e5" Time="1962" Action="horrible" Param="childhood" />
<Event Id="e6" Time="1963" Action="best friend" Param="i151" />
<Event Id="e7" Time="1964" Action="believe" Param="god" />
<Event Id="e8" Time="1964" Action="every week go" Param="church" />
...
<Event Id="e16" Time="1968" Action="problems" Param="drugs" />
<Event Id="e17" Time="1971" Action="grow" Param="adult" />
<Event Id="e18" Time="1971" Action="friend" Param="i98" />
<Event Id="e19" Time="1972" Action="involved" Param="labour union" />
<Event Id="e20" Time="1972" Action="friend" Param="i156" />
<Event Id="e21" Time="1973" Action="get" Param="arrested" />
<Event Id="e22" Time="1973" Action="learn" Param="play guitar" />
<Event Id="e23" Time="1975" Action="became" Param="hippy" />
...
<Event Id="e36" Time="1985" Action="divorce" Param="i439" />
<Event Id="e37" Time="1987" Action="couple" Param="i102" />
<Event Id="e38" Time="1987" Action="live together" Param="i102" />
<Event Id="e39" Time="1987" Action="have" Param="abortion" />
...
</Log>
<Log Id="i50">
<Description />
<Attribute Id="name" Value=“francisco" />
...
Samer Hassan
SSASA 2008
27
An example: part of the life-story generated





Rosa Pérez was born in 1955, and she met Luis Martínez, and she met
Miguel López. She suffered a horrible childhood, and she had a very
good friend: María Valdés, and she believed in God, and she used to go
to church every week.
. . .
When she was a teenager, (...) she had problems with drugs, and she
became an adult, and she met Marci Boyle, and while she was involved
in a labour union, she met Carla González and she got arrested. She
learned how to play the guitar, and so she became a hippy, getting
involved in a NGO.
. . .
She met Sara Hernández, and she stopped going to church, and she met
Marcos Torres, and she fell in love, desperately, with Marcos Torres,
but in the end she went out with Miguel López, and she co-habitated
with Miguel López, and she had a child: Melvin López.
. . .
She met Sergio Ruiz, and she separated from Miguel López, and she went
out with Sergio Ruiz, and she co-habitated with Sergio Ruiz. She had a
abortion, and so she had a depression, and she had a crisis of values.
She was unfaithful to Sergio Ruiz with another man.
. . .
Nowadays she is an atheist.
Samer Hassan
SSASA 2008
28
Contents

The Problem

ABM Mentat: Design

ABM Mentat: Results

AI: Fuzzy Logic

AI: Natural Language Processing

AI: Data Mining
Samer Hassan
SSASA 2008
29
Introduction of AI: Data Mining

Data Mining is the process of extracting patterns and
relevant information from large amounts of data

Design:


Pre-processing of empirical data (surveys):


Allows simplification, locates redundant attributes
Clustering: selection of qualitative “ideal types”
Post-processing of simulation output:

Clustering:
• Shows non-visible patterns
• Comparison of patterns
• Different life-stories for each pattern

Samer Hassan
Classification: evolution of “ideal types”
SSASA 2008
30
Limitations & Future Work

Enough demography!


Quest for a proper cognitive model for this task



Overcome methodological limitation: implementing
diffusion of moral values
...or forget about it
definitely not BDI
Improve other aspects:



Samer Hassan
ABM design (Ex: friendship ties may weaken)
Fuzzy inference
Quality of biographies
SSASA 2008
31
Thanks for your attention!
Samer Hassan
[email protected]
Samer Hassan
SSASA 2008
32
Contents License

This presentation is licensed under a
Creative Commons Attribution 3.0
http://creativecommons.org/licenses/by/3.0/
Para ver esta película, debe
disponer de QuickTime™ y de
un descompresor TIFF (sin comprimir).

You are free to copy, modify and distribute it as long as
the original work and author are cited
Samer Hassan
SSASA 2008
33