[PDF] - Emergent systems Spring-14 Cultural models PDF Document

SLIDE 1

1

Umeå University Department of Computing Science

Emergent systems

Spring-14 Cultural models

http://www.cs.umu.se/kurser/5DV017

Emergent Systems, Jonny Pettersson, UmU 10/2 - 14

Previous lectures

❒ Self-Organization ❒ Autonomous Agents ❒ Real Ants ❒ Virtual Ants ❒ Ant Algorithms ❒ Schooling of fish ❒ Boids ❒ Assignment 2

Emergent Systems, Jonny Pettersson, UmU 10/2 - 14

Outline for today

❒ Swarm algorithms ❒ Culture models

SLIDE 2

2

Emergent Systems, Jonny Pettersson, UmU 10/2 - 14

Swarm Algorithms

❒ Solve problems using cooperative behavior

f a group of agents

❒ Santa Fe

❍ ”We use the term ’swarm’ in a general sense to

refer to any loosely structured collection of interacting agents.” ❒ Each agent is only part of the overall

solution

❍ Termites ❍ Boids ❍ Ant algorithms ❍ ...

Emergent Systems, Jonny Pettersson, UmU 10/2 - 14

Culture Models

❒ Culture

❍ ”[A term] used to indicate the set of individual

attributes that are subject to social influence.” (Axelrod, The Dissemination of Culture, 1997) ❒ Models

❍ Axelrod’s Culture Model ❍ The Party Model ❍ The Segregation Model ❍ (The last two models is inspired by Thomas

Schelling's writings about social systems)

Emergent Systems, Jonny Pettersson, UmU

Axelrod, The Culture Model, 97

❒ Question:

❍ If people tend to become more alike in their

beliefs, attitudes and behavior when they interact, why do not all such differences eventually disappear? ❒ To understand how a culture can get

established, spread and be sustained has great importance

❍ State formation, succession conflicts,

transnational integration, domestic cleavages ❒ Many explanations has been suggested

10/2 - 14

SLIDE 3

3

Emergent Systems, Jonny Pettersson, UmU

The model - Assumptions

❒ Based on two assumptions:

❍ Individuals interact with higher possibilities

with others that share many of their cultural features

❍ Interaction between two individuals tends to

increase the number of shared cultural features ❒ Similarity è Interaction è ì Similarity

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

The model - Principles

❒ Based on three principles:

❍ Agent based modeling ❍ No central authority ❍ Adaptive agents

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

The model - Representation

❒ An individuals culture is represented by a

list of features

❒ Each feature have one trait of a set of

traits specific for that feature

❒ Ex. 5 features with 10 traits/feature 34812 79911 85312 91237 ❒ Cultural similarity = ♯ shared features /

♯ features

❒ A 2D grid, 4 neighbors, no wrap around

10/2 - 14

SLIDE 4

4

Emergent Systems, Jonny Pettersson, UmU

The Algorithm

❒ Repeat the following steps for as many

events as desired

1.

At random, pick one site to be active

2. At random, pick one of the neighbors to the

active site

3. With probability equal to their cultural

similarity these two sites interact

4. If they interact, at random select one of the

feature that has different traits. Copy the neighbors trait to the active site ❒ The simulation is done one event at a time

to avoid synchronization

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

Definitions

❒ Cultural region

❍ A region with a set of contiguous sites with an

identical culture ❒ Stable region

❍ When a cultural region has nothing in common

with any adjacent regions

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

Number of Stable Regions

❒ Depends on:

❍ The number of features ❍ The number of traits/feature ❍ The range of interactions ❍ The size of the geographic territory

10/2 - 14

SLIDE 5

5

Emergent Systems, Jonny Pettersson, UmU

Hypotheses 1 and 2

1. More features è More stable regions
2. More traits/feature è More stable

regions

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

Experiment

1. More features è More stable regions

❍

Not correct

2. More traits/feature è More stable

regions

❍

Correct

10/2 - 14

Size: 10x10

Traits/feature 5 10 15 Features 5 1.0 3.2 20.0 10 1.0 1.0 1.4 15 1.0 1.0 1.0

Emergent Systems, Jonny Pettersson, UmU

Hypothesis 3 – Range of Interactions

❒ Greater range of interactions è Less

stable regions

❒ Hypothesis 3 is correct

10/2 - 14

Number of neighbors Number of stable regions 4 3.4 8 2.5 12 1.5

Average of 9 types of cultures, 10 times/type of culture

SLIDE 6

6

Emergent Systems, Jonny Pettersson, UmU

Hypothesis 4 – Size of the Territory

❒ More sites è More stable regions ❒ Fig 2

❍ 5 features, 15 traits/feature, 4 neighbors

❒ Hypothesis 4 is not correct ❒ Why do large territories have fewer stable

regions than moderate-sized territories?

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

Explanation

❒ Random walk with absorbing barriers

111 111 111 111 112 112

í î

111 111 111 111 111 112 111 111 111 112 112 112

è The larger region “eat” the smaller

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

Explanation (cont.)

❒ Cultural zones

❍ A set of contiguous sites, each of which has a

neighbor they can interact with

❍ A run ends when a zone has exactly one region ❍ Fig 3

Compatible cultures “struggles for survival

10/2 - 14

SLIDE 7

7

Emergent Systems, Jonny Pettersson, UmU

Explanation (cont.)

❒ Over time, boundaries between regions in

the same cultural zone tend to dissolve

❍ On average, cultural similarity between

adjacent sites in the same cultural zone tends to increase

❍ Even boundaries between cultural zones can

dissolve

❍ Traits move around in a zone, the longer the

time, the higher probability to dissolve a boundary

❍ A doubling of the number of sites, allows four

times as many activations in all

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

Explanation - Summary

❒ Why larger territories have fewer stable

regions than moderate-sized territories

❍ In small territories, there is not room for many

stable regions

❍ In moderate-sized territories, there is enough

sites

❍ In large territories, there is even more room,

but the process of social influence and the consequent movement of cultural alternatives go on so long that virtually all cultural boundaries eventually dissolve

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

Conclusions

❒ Intuition is not always right ❒ Local convergence, can lead to global

polarization

10/2 - 14

SLIDE 8

8

Emergent Systems, Jonny Pettersson, UmU 10/2 - 14

Culture Models

❒ Culture

❍ ”[A term] used to indicate the set of individual

attributes that are subject to social influence.” (Axelrod, The Dissemination of Culture, 1997) ❒ Models

❍ Axelrod’s Culture Model ❍ The Party Model ❍ The Segregation Model ❍ (The last two models is inspired by Thomas

Schelling's writings about social systems)

Emergent Systems, Jonny Pettersson, UmU

Scientific approach

23/1 - 14

Formulate the question Communicate the model Assemble hypotheses Choose model structure Implement the model Analyze the model Patterns Patterns

(Adapted from Grimm and Railsback 2005)

Emergent Systems, Jonny Pettersson, UmU 23/1 - 14

Assignment 1 – General feedback

❒ The report should have a scientific

approach

❍ Introduction/Background ❍ Method ❍ Result ❍ Discussion

SLIDE 9

9

Emergent Systems, Jonny Pettersson, UmU

Assignment 1 - Reflections

❒ How did you validate your method for deciding

nly one pile?

❒ How did you validate your results? ❒ How many repetitions for each parameter

settings did you do and why?

❒ Is your report freestanding from the

specification of the assignment?

❒ Can your results be reproduced with only the

information in your report?

❒ Did you clearly answer all the questions? ❒ Did you clearly followed up all your

hypotheses?

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU

Assignment 1 - Questions

❒ How many termites is needed? ❒ Density? ❒ Efficiency of the system? ❒ Applications for systems like this?

10/2 - 14 Emergent Systems, Jonny Pettersson, UmU 10/2 - 14

Summary

❒ Swarm algorithms ❒ Culture models

❍ Axelrod’s Culture Model ❍ The Party Model ❍ The Segregation Model