Uncertainty, cooperation, communication complexity, and social network structure Peter Andras

Slides:



Advertisements
Similar presentations
Approaches, Tools, and Applications Islam A. El-Shaarawy Shoubra Faculty of Eng.
Advertisements

Agent-based Modeling: A Brief Introduction Louis J. Gross The Institute for Environmental Modeling Departments of Ecology and Evolutionary Biology and.
Week 11 Review: Statistical Model A statistical model for some data is a set of distributions, one of which corresponds to the true unknown distribution.
Evolution of Cooperation The importance of being suspicious.
PROBABILITY. Uncertainty  Let action A t = leave for airport t minutes before flight from Logan Airport  Will A t get me there on time ? Problems :
VK Dice By: Kenny Gutierrez, Vyvy Pham Mentors: Sarah Eichhorn, Robert Campbell.
Linear Transformation and Statistical Estimation and the Law of Large Numbers Target Goal: I can describe the effects of transforming a random variable.
©GoldSim Technology Group LLC., 2004 Probabilistic Simulation “Uncertainty is a sign of humility, and humility is just the ability or the willingness to.
Minority Games A Complex Systems Project. Going to a concert… But which night to pick? Friday or Saturday? You want to go on the night with the least.
Tags and Image Scoring for Robust Cooperation By Nathan Griffiths Presented at AAMAS 2008.
A Memetic Framework for Describing and Simulating Spatial Prisoner’s Dilemma with Coalition Formation Sneak Review by Udara Weerakoon.
Decision Analysis April 11, Game Theory Frame Work Players ◦ Decision maker: optimizing agent ◦ Opponent  Nature: offers uncertain outcome  Competition:
The Evolution of Cooperation Shade Shutters School of Life Sciences & Center for Environmental Studies.
Evolution of Complex Systems Lecture 3: Theoretical foundations Peter Andras / Bruce Charlton
Evolutionary Computation Introduction Peter Andras s.
Human Social Dilemmas Cooperation Between Non-Relatives Complex Evolutionary Problem Repeated Interaction, Conditional Cooperation Human Cooperation Often.
458 Age-structured models (Individual-based versions) Fish 458, Lecture 6.
Altruism A Simulated Investigation COM SCI 194 Honors Research Fall 2007 ~ Spring 2008 Alexander Liu and Eric Chang Professor Amit Sahai.
Management from the perspective of systems theory Peter Andras and Bruce G Charlton University of Newcastle
Evolutionary Computation Application Peter Andras peter.andras/lectures.
On Bounded Rationality and Computational Complexity Christos Papadimitriou and Mihallis Yannakakis.
Slide 1 Statistics Workshop Tutorial 4 Probability Probability Distributions.
Artificial Neural Networks -Application- Peter Andras
STOCHASTIC GEOMETRY AND RANDOM GRAPHS FOR THE ANALYSIS AND DESIGN OF WIRELESS NETWORKS Haenggi et al EE 360 : 19 th February 2014.
Non verbal Communication. What is non verbal communication?
Large-scale organization of metabolic networks Jeong et al. CS 466 Saurabh Sinha.
Agent Based Modeling and Simulation
Conference Paper by: Bikramjit Banerjee University of Southern Mississippi From the Proceedings of the Twenty-Seventh AAAI Conference on Artificial Intelligence.
5. Alternative Approaches. Strategic Bahavior in Business and Econ 1. Introduction 2. Individual Decision Making 3. Basic Topics in Game Theory 4. The.
Experiments & Statistics. Experiment Design Playtesting Experiments don’t have to be “big”--many game design experiments take only 30 minutes to design.
P systems: A Modelling Language Marian Gheorghe Department of Computer Science University of Sheffield Unconventional Programming Paradigms; Sept’04.
Estimation of Statistical Parameters
Learning in Multiagent systems
K. J. O’Hara AMRS: Behavior Recognition and Opponent Modeling Oct Behavior Recognition and Opponent Modeling in Autonomous Multi-Robot Systems.
Linking multi-agent simulation to experiments in economy Re-implementing John Duffy’s model of speculative learning agents.
Example Department of Computer Science University of Bologna Italy ( Decentralised, Evolving, Large-scale Information Systems (DELIS)
Rationality meets the tribe: Some models of cultural group selection David Hales, The Open University Hales, D., (2010) Rationality.
The Evolution of Fairness PSC 120 Jeff Schank. Fairness People engage in fair exchanges of resources even when it would benefit them more to act unfairly.
Boltzmann Machine (BM) (§6.4) Hopfield model + hidden nodes + simulated annealing BM Architecture –a set of visible nodes: nodes can be accessed from outside.
Aemen Lodhi (Georgia Tech) Amogh Dhamdhere (CAIDA)
Ch 6-1 © 2004 Pearson Education, Inc. Pearson Prentice Hall, Pearson Education, Upper Saddle River, NJ Ostwald and McLaren / Cost Analysis and Estimating.
Decision making Under Risk & Uncertainty. PAWAN MADUSHANKA MADUSHAN WIJEMANNA.
1 Chapter 8 Sampling Distributions of a Sample Mean Section 2.
Project funded by the Future and Emerging Technologies arm of the IST Programme FET-Open scheme Project funded by the Future and Emerging Technologies.
Integral projection models
Game Theory, Social Interactions and Artificial Intelligence Supervisor: Philip Sterne Supervisee: John Richter.
Evolving cooperation in one-time interactions with strangers Tags produce cooperation in the single round prisoner’s dilemma and it’s.
Monte Carlo Process Risk Analysis for Water Resources Planning and Management Institute for Water Resources 2008.
Environmental and Societal Impacts Group Emergency Management Higher Education Conference Seth McGinnis
Changing the Rules of the Game Dr. Marco A. Janssen Department of Spatial Economics.
Genome Biology and Biotechnology The next frontier: Systems biology Prof. M. Zabeau Department of Plant Systems Biology Flanders Interuniversity Institute.
Evolving Specialisation, Altruism & Group-Level Optimisation Using Tags David Hales Centre for Policy Modelling, Manchester Metropolitan University, UK.
Simulation. Types of simulation Discrete-event simulation – Used for modeling of a system as it evolves over time by a representation in which the state.
Evolution of Cooperation in Mobile Ad Hoc Networks Jeff Hudack (working with some Italian guy)
Building Valid, Credible & Appropriately Detailed Simulation Models
Surface Defect Inspection: an Artificial Immune Approach Dr. Hong Zheng and Dr. Saeid Nahavandi School of Engineering and Technology.
Statistical Concepts Basic Principles An Overview of Today’s Class What: Inductive inference on characterizing a population Why : How will doing this allow.
Expected Return and Risk. Explain how expected return and risk for securities are determined. Explain how expected return and risk for portfolios are.
Social Networks and Trust: not the Experimental Evidence you may Expect Daniela Di Cagno Emanuela Sciubba Luiss Guido Carli, Rome Birkbeck College, London.
Analysis…Measures of Central Tendency How can we make SENSE of our research data???
Modelling and Simulating Social Systems with MATLAB
Environmental and Societal Impacts Group Emergency Management Higher Education Conference Seth McGinnis.
Modelling and Simulating Social Systems with MATLAB
Context dependent pollinator limitation
By C. Kohn Waterford Agricultural Sciences
Additional notes on random variables
CHAPTER 15 SUMMARY Chapter Specifics
Additional notes on random variables
Boltzmann Machine (BM) (§6.4)
Psych 231: Research Methods in Psychology
Presentation transcript:

Uncertainty, cooperation, communication complexity, and social network structure Peter Andras

2 Overview Uncertainty and cooperation Uncertainty and cooperation Components of uncertainty Components of uncertainty Communication of intentions Communication of intentions Uncertainty and communication complexity Uncertainty and communication complexity Artificial social network structure Artificial social network structure

3 Cooperation Individuals are selfish – Why do they cooperate ? Individuals are selfish – Why do they cooperate ? Examples: microbes, worms, ants, fish, wolves, humans Examples: microbes, worms, ants, fish, wolves, humans

4 Theories of cooperation Theories: Theories: reciprocity (direct/indirect) reciprocity (direct/indirect) similarity (tags/genes/etc) similarity (tags/genes/etc) commitment inertia commitment inertia assortment/segregation assortment/segregation Methods of study: Methods of study: live experiments (bacteria, plants, animals, humans) live experiments (bacteria, plants, animals, humans) agent-based simulations agent-based simulations

5 Uncertainty Sources of uncertainty: predators, food scarcity, extreme natural conditions (cold, hot, wet, dry) Sources of uncertainty: predators, food scarcity, extreme natural conditions (cold, hot, wet, dry) Usually more cooperation in uncertain environments Usually more cooperation in uncertain environments alpine plants alpine plants microbes in presence of antibiotics microbes in presence of antibiotics mole-rats in dry environment mole-rats in dry environment fish in high predation risk environment fish in high predation risk environment humans during natural disasters or wars humans during natural disasters or wars

6 Agent-based simulation Agents play cooperation games (e.g. Prisoner’s Dilemma) with other agents Agents play cooperation games (e.g. Prisoner’s Dilemma) with other agents Usually: many repeated games with all possible partners Usually: many repeated games with all possible partners

7 Simulation of uncertainty Usual game matrix Usual game matrix Game matrix with uncertainty Game matrix with uncertainty P=N(p,  ), Q=N(q,  ), R=N(r,  ), S=N(s,  ) P=N(p,  ), Q=N(q,  ), R=N(r,  ), S=N(s,  ) Player 1 Player 2 CpDf Cpp,pr,q Dfq,rs,s Player 1 Player 2 CpDf CpP,PR,Q DfQ,RS,S

8 Agent-based simulation Our simulation: Our simulation: moving agents in 2D world moving agents in 2D world random selection of interaction partners from neighborhood random selection of interaction partners from neighborhood finite life finite life offspring generation offspring generation resource accumulation and consumption resource accumulation and consumption resource generation through game playing resource generation through game playing

9 Uncertainty and cooperation more uncertainty  more cooperation more uncertainty  more cooperation (Andras et al, 2003, in: Adaptive Agents and Multi-Agent Systems, pp.49-65; Andras et al, 2007, BMC Evolutionary Biology, 7:240)

10 Representation of uncertainty Uncertainty is present in natural environments of living organisms in the form of the variance of outcomes of events or scenarios involving the organism Uncertainty is present in natural environments of living organisms in the form of the variance of outcomes of events or scenarios involving the organism Representation: variance of resources Representation: variance of resources

11 Objective uncertainty Suppose the environment is described in terms of resources Suppose the environment is described in terms of resources Objective uncertainty is the variance of the resource distribution Objective uncertainty is the variance of the resource distribution

12 Subjective uncertainty The observable range of resources is different from the natural range of them The observable range of resources is different from the natural range of them E.g., resource amounts too little to be worth exploring E.g., resource amounts too little to be worth exploring

13 Subjective  2 > Objective  2 Subjective uncertainty is higher than the objective uncertainty, given that at least half of the natural resource distribution is in the observable range Subjective uncertainty is higher than the objective uncertainty, given that at least half of the natural resource distribution is in the observable range

14 Effective uncertainty Individuals share their subjective uncertainty through cooperation. The experienced uncertainty is the effective uncertainty. Individuals share their subjective uncertainty through cooperation. The experienced uncertainty is the effective uncertainty.

15 Effective  2 < Subjective  2 Effective uncertainty is smaller than subjective uncertainty if the individuals cooperate Effective uncertainty is smaller than subjective uncertainty if the individuals cooperate

16 Steady-state uncertainty Effective uncertainty is reduced through cooperation to the level of steady-state uncertainty that allows reproduction or stable growth of the agent population Effective uncertainty is reduced through cooperation to the level of steady-state uncertainty that allows reproduction or stable growth of the agent population (Andras et al, 2006, JASSS – Journal of Artificial Societies and Social Simulation, 9:1/7)

17 Communication of intentions Organisms communicate with other organisms about their intentions – this plays an important role in cooperation and cheating Organisms communicate with other organisms about their intentions – this plays an important role in cooperation and cheating E.g. exposure of signal molecules on the cell surface, vocalisations and postures of animals, gestures, body language and spoken language of humans E.g. exposure of signal molecules on the cell surface, vocalisations and postures of animals, gestures, body language and spoken language of humans

18 Agent language – 1 Intention of cooperation – I coop Intention of cooperation – I coop Language = lexicon, syntax, semantics lexicon = {0,s,i,y,n,h,t} syntax = probabilistic two-input automaton – –E.g. s,i’  0.6 i;  0.3 y;  0.1 n semantics = 0 – no interest, s – start of communication, i – intend to communicate further, y – want to engage in cooperation, n – lost interest, h – cooperate, t – defect P(h|y,y’) = I coop P(h|y,y’) = I coop (Andras et al, 2003, in: Adaptive Agents and Multi-Agent Systems, pp.49-65; Andras, 2008a, in: Proceedings of the IEEE Conference on Evolutionary Computation; Andras, 2008b, in: Proceedings of the Artificial Life XI)

19 Agent language – 2 Intention consistency: a big smile is more likely to follow a small smile than an angry face Positivity order: n, 0, s, i, y Intention consistency rules: P(x1,x’  y)  P(x2,x’  y) if Pos(y)  Pos(x1)  Pos(x2) P(x,x1’  y)  P(x,x2’  y) if Pos(y)  Pos(x) & Pos(x1)  Pos(x2)

20 Uncertainty and language Ambiguous use of the language may add to the uncertainty induced by other environmental factors Ambiguous use of the language may add to the uncertainty induced by other environmental factors High uncertainty may lead to lower ambiguity of the language High uncertainty may lead to lower ambiguity of the language army army surgical theatre surgical theatre West-African languages West-African languages Ambiguity ~ lexical complexity Ambiguity ~ lexical complexity

21 Language complexity Kolmogorov complexity – description length measure Kolmogorov complexity – description length measure Variance of transition probabilities (e.g. P(y|i,i’)) – variability of language usage Variance of transition probabilities (e.g. P(y|i,i’)) – variability of language usage Lexical language complexity: average of transition probability variances Lexical language complexity: average of transition probability variances

22 Uncertainty and language complexity more uncertainty  less lexical language complexity more uncertainty  less lexical language complexity (Andras, 2008b, in: Proceedings of the Artificial Life XI)

Artificial social networks Can we generate agent-based simulations of social interaction systems that have scale-free interaction networks ? (without explicitly encoding to have this interaction network in the simulation) Does the presence of memories, gossip and uncertainty in the simulation matter for this? 23

Memory and gossip Memory: the agents remember their interactions with other agents and accordingly adapt their willingness to cooperate Memory: the agents remember their interactions with other agents and accordingly adapt their willingness to cooperate Gossip: the agents share their memories about other agents with their interaction partners Gossip: the agents share their memories about other agents with their interaction partners 24

Social network measurement simulations for each setting running for 1000 turns, measurement for consecutive 100 turns 20 simulations for each setting running for 1000 turns, measurement for consecutive 100 turns Settings: Settings: low / high uncertainty low / high uncertainty with / without memory with / without memory with / without gossip (in case of with memory) with / without gossip (in case of with memory) Measurement of the interaction network Measurement of the interaction network Expected connectedness distribution: Expected connectedness distribution: Exponent estimated as: Exponent estimated as:

Results Kolmogorov-Smirnov test (Matlab) was used to check the match between measured and expected distributions Kolmogorov-Smirnov test (Matlab) was used to check the match between measured and expected distributions The log(p) is the logarithm of the calculated significance level – the network is significantly different from a scale-free network if log(p) < –2. The log(p) is the logarithm of the calculated significance level – the network is significantly different from a scale-free network if log(p) < –2. 26

27

Results The memory and gossip settings have no significant effect on the power law nature of the connectedness distribution of the corresponding simulated social networks The memory and gossip settings have no significant effect on the power law nature of the connectedness distribution of the corresponding simulated social networks The presence of uncertainty is critical for the generation of interaction networks with scale- free connectedness distribution The presence of uncertainty is critical for the generation of interaction networks with scale- free connectedness distribution 28

29 Summary More uncertain environments induce more cooperation More uncertain environments induce more cooperation Uncertainty: objective, subjective effective uncertainty; Objective  2 < Subjective  2, Effective  2 < Subjective  2 Uncertainty: objective, subjective effective uncertainty; Objective  2 < Subjective  2, Effective  2 < Subjective  2 More uncertainty induces reduction of the lexical complexity of the language used to communicate intentions More uncertainty induces reduction of the lexical complexity of the language used to communicate intentions Artificial social networks are more similar to natural ones in presence of uncertainty, gossip and memory does not seem to have an impact on this Artificial social networks are more similar to natural ones in presence of uncertainty, gossip and memory does not seem to have an impact on this

30 Acknowledgement John Lazarus John Lazarus Gilbert Roberts Gilbert Roberts