Annexations and Merging in Weighted Voting Games: The Extent of Susceptibility of Power Indices by Ramoni Lasisi Vicki Allan.

Slides:



Advertisements
Similar presentations
Chapter 2: Weighted Voting Systems
Advertisements

Weighted Voting, Algorithms and Voting Power
Weighted Voting When we try to make collective decisions, it is only natural to consider how things are done in society. We are familiar with voting for.
Chapter 11: Weighted Voting Systems Lesson Plan
Presented by: Katherine Goulde
Chapter 13 – Weighted Voting Part 4 Appropriate applications of measures of power Minimal winning coalitions Classification of weighted voting systems.
Manipulation, Control, and Beyond: Computational Issues in Weighted Voting Games Edith Elkind (U. of Southampton) Based on joint work with: Y.Bachrach,
Effort Games and the Price of Myopia Michael Zuckerman Joint work with Yoram Bachrach and Jeff Rosenschein.
1 Message to the user... The most effective way to use a PowerPoint slide show is to go to “SLIDE SHOW” on the top of the toolbar, and choose “VIEW SHOW”
Excursions in Modern Mathematics Sixth Edition
Yoram Bachrach Jeffrey S. Rosenschein November 2007.
Manipulation and Control in Weighted Voting Games Based on: Bachrach, Elkind, AAMAS’08 Zuckerman, Faliszewski, Bachrach, Elkind, AAAI’08.
Homework Discussion Read Pages 48 – 62 Page 72: 1 – 4, 6 TEST 1 ON THURSDAY FEBRUARY 8 –The test will cover sections 1.1 – 1.6, and 2.1 – 2.3 in the textbook.
Path Disruption Games (Cooperative Game Theory meets Network Security) Yoram Bachrach, Ely Porat Microsoft Research Cambridge.
Andrea Katz April 29, 2004 Advisor: Dr. Karrolyne Fogel Voting Blocs in Academic Divisions of CLU: A Mathematical Explanation of Faculty Power.
Chapter 8 Two-Level Fractional Factorial Designs
Ensemble Learning: An Introduction
CHAPTER 7: SAMPLING DISTRIBUTIONS. 2 POPULATION AND SAMPLING DISTRIBUTIONS Population Distribution Sampling Distribution.
Computing the Banzhaf Power Index in Network Flow Games
Approximating Power Indices Yoram Bachrach(Hebew University) Evangelos Markakis(CWI) Ariel D. Procaccia (Hebrew University) Jeffrey S. Rosenschein (Hebrew.
Yair Zick, Alexander Skopalik and Edith Elkind School of Physical and Mathematical Sciences Division of Mathematical Sciences Nanyang Technological University,
The Agencies Method for Coalition Formation in Experimental Games John Nash (University of Princeton) Rosemarie Nagel (Universitat Pompeu Fabra, ICREA,
Excursions in Modern Mathematics, 7e: Copyright © 2010 Pearson Education, Inc. 2 The Mathematics of Power 2.1An Introduction to Weighted Voting.
Sample Of size 2 Of size 3 1 A,B=3,1 2 A,B,C=3,1,5 3 A,C=3,5 4
JCKBSE2010 Kaunas Predicting Combinatorial Protein-Protein Interactions from Protein Expression Data Based on Correlation Coefficient Sho Murakami, Takuya.
A SEARCH-BASED APPROACH TO ANNEXATION AND MERGING IN WEIGHTED VOTING GAMES Ramoni Lasisi and Vicki Allan Utah State University by.
Calculating Power in Larger Systems Nominal, Banzhaf and Shapley-Shubik.
Chapter 13 – Weighted Voting
Chapter 13: Weighted Voting Banzhaf Power Index Shapley-Shubik Power Index Equivalent Systems Examples.
Weighted Voting Systems Brian Carrico. What is a weighted voting system?  A weighted voting system is a decision making procedure in which the participants.
Statistical Analysis Professor Lynne Stokes Department of Statistical Science Lecture 14 Sequential Experimentation, Screening Designs, Fold-Over Designs.
A small country has 4 provinces – A, B, C and D. Each province contains 30%, 20%, 10% and 40% of the population in the country, respectively. The.
Fractional Factorial Design Full Factorial Disadvantages Full Factorial Disadvantages –Costly (Degrees of freedom wasted on estimating higher order terms)
Statistics for Managers Using Microsoft Excel, 5e © 2008 Pearson Prentice-Hall, Inc.Chap 8-1 Statistics for Managers Using Microsoft® Excel 5th Edition.
Math for Liberal Studies.  We want to measure the influence each voter has  As we have seen, the number of votes you have doesn’t always reflect how.
Weighted Voting Systems
Slide Slide 1 Copyright © 2007 Pearson Education, Inc Publishing as Pearson Addison-Wesley. Overview.
Manipulating the Quota in Weighted Voting Games (M. Zuckerman, P. Faliszewski, Y. Bachrach, and E. Elkind) ‏ Presented by: Sen Li Software Technologies.
1 What is Game Theory About? r Analysis of situations where conflict of interests is present r Goal is to prescribe how conflicts can be resolved 2 2 r.
Theorem: Equal weight implies equal power but not the converse.
CONSENSUS THEOREM Choopan Rattanapoka.
An importer of Herbs and Spices claims that average weight of packets of Saffron is 20 grams. However packets are actually filled to an average weight,
An importer of Herbs and Spices claims that average weight of packets of Saffron is 20 grams. However packets are actually filled to an average weight,
Section 1.1, Slide 1 Copyright © 2014, 2010, 2007 Pearson Education, Inc. Section 11.3, Slide 1 11 Voting Using Mathematics to Make Choices.
Data Mining Association Rules Mining Frequent Itemset Mining Support and Confidence Apriori Approach.
Chapter 11. Weighted Voting Systems  Goals Study weighted voting systems ○ Coalitions ○ Dummies and dictators ○ Veto power Study the Banzhaf power index.
Weighted Voting Systems Chapter 2 Objective: Calculate the Banzhaf power Index for a weighted voting system. Learn additional notation and terminology.
Election Theory A Tale of Two Countries or Voting Power Comes To the Rescue!
Review Homework Pages , 46, 56 2.Combination – order is not important 3.Permutation – order is important 4. 5.ABC, ABD, ABE, ACD, ACE, ADE,
The Shapley Value The concept of the core is useful as a measure of stability. As a solution concept, it presents a set of imputations without distinguishing.
Non-transitivity and Probability Steven Kumer Advised by Dr. Bryan Shader.
1 EFFICIENCY OF FAIRNESS IN VOTING SYSTEMS EPCS 2009 Athens, April 2-5, 2009 Frantisek Turnovec Charles University in Prague Institute of Economic Studies.
Dr. Vicki Allan Multiagent systems – program computer agents to act for people. If two heads are better than one, how about 2000?
Dynamic Weighted Voting Games Edith Elkind Dmitrii Pasechnik Yair Zick AAMAS 2013.
EXCURSIONS IN MODERN MATHEMATICS SIXTH EDITION Peter Tannenbaum 1.
The Banzhaf Power Index
Excursions in Modern Mathematics Sixth Edition
Chapter 11: Weighted Voting Systems Lesson Plan
Lesson 2.9 Objective: Probability permutations and combinations
Math 210G Mathematics Appreciation Su Voto es Su Voz
Excursions in Modern Mathematics Sixth Edition
Fractional Factorial Design
Confidence Intervals for a Standard Deviation
Exploring Partially ordered sets
Homework Collection Homework Discussion Page 35: 35, 38, 63, 64
Chapter 11: Weighted Voting Systems Lesson Plan
The Banzhaf Power Index
Discrete Math Weighted Voting.
Design matrix Run A B C D E
FAIRNESS AND EFFICIENCY IN VOTING SYSTEMS
Presentation transcript:

Annexations and Merging in Weighted Voting Games: The Extent of Susceptibility of Power Indices by Ramoni Lasisi Vicki Allan

Agenda Weighted Voting Games (WVGs) Power Indices : Shapley-Shubik, Banzhaf, & Deegan-Packel Manipulation of WVGs : Annexations & Merging

WVGs - mathematical abstractions of voting systems. Let V be the set of voters. A weight function is defined on V, w: V Q +. A coalition of agents C, wins in the game if the sum of their weights meets or exceeds a threshold called the quota q. C is also called a winning coalition. Representation of WVG: G = [w 1,w 2,…, w n ; q] Weighted Voting Games (WVGs)

Example Banzhaf wanted to prove the Nassau County board’s voting system (based on population) was unfair – Hempstead #1: 9 – Hempstead #2: 9 – North Hempstead: 7 – Oyster Bay: 3 – Glen Cove: 1 – Long Beach: 1 This is 30 total votes, and a simple majority of 16 votes was required for a measure to pass.

[9, 9, 7, 3, 1, 1 :16 ] Look at all possible winning coalitions AB AC BC ABC ABD ABE ABF ACD ACE ACF BCD BCE BCF ABCD ABCE ABCF ABDE ABDF ABEF ACDE ACDF ACEF BCDE BCDF BCEF ABCDE ABCDF ABCEF ABDEF ACDEF BCDEF ABCDEF Determine which voters are CRITICAL (underlined) to each coalition. Critical if not have enough votes without voter. AB AC BC ABC ABD ABE ABF ACD ACE ACF BCD BCE BCF ABCD ABCE ABCF ABDE ABDF ABEF ACDE ACDF ACEF BCDE BCDF BCEF ABCDE ABCDF ABCEF ABDEF ACDEF BCDEF ABCDEF

Look at the number of swing votes 48 swing votes Determine what proportion of the swing votes are held by each voter All power indices are a probability. [9, 9, 7, 3, 1, 1 :16 ] – Hempstead #1 = 16/48 – Hempstead #2 = 16/48 – North Hempstead = 16/48 – Oyster Bay = 0/48 – Glen Cove = 0/48 – Long Beach = 0/48 Banzhaf argued that a voting arrangement that gives 0% of the power to 16% of the population is unfair.

How do we evaluate the strength of agents in WVGs? Using Power Indices Power Indices a fraction of the power attributed to each voter Power: Not proportional to voting weight Your ability to change the outcome with your vote The probability of having a significant role in determining the outcome There are different definitions of ‘having a significant role’

Various Power indices Banzhaf index: The number of winning coalitions in which an agent is critical. Denoted by ß i (G). Critical – swing agent in a winning coalition is an agent that causes the coalition to lose when removed from it. Considers all the marginal contributions of a player to all possible coalitions, without considering the order of the players Normalize so add to one The power index is the portion of coalitions in which the agent is critical

A B Quota Banzhaf Power Index A A B C C [4,2,3:6] What is critical?

A B Quota Banzhaf Power Index A A B C C There are three winning coalitions -A is critical in all three -B is critical in only one -C is critical in only one 5 total swing votes A = 3/( ); B = C = 1/( ) [4,2,3:6]

Banzhaf Power Distribution A B C [4,2,3:6]

Various Power indices Shapley-Shubik index: Value added. The number of permutations of the set of agents for which an agent is critical. Denoted by φ i (G). – What would you be without me? – Well, what would YOU be without me? – Solution – consider all possible orders.

A B C Quota How important is each voter? Shapley Shubik Look at value added. What do I add to the existing group? Consider the group being formed one at a time. [4,2,3: 6]

A B C Quota How important is each voter? A A A A A B B B B B C C C C C

A claims 2/3 of the power in this scheme

A B C Quota If quota changes, power shift. A A A A A B B B B B C C C C C

Now, power is equal!

Various Power indices Deegan-Packel index: Depends on the number of Minimal Winning Coalitions (MWCs). Within each winning coalition, the credit is shared equally (as if any is removed, the coalition fails) Thus, the size of each of the MWCs that include the agent is considered. Denoted by γ i (G).

A Quota Deegan-Packel Power Index A B C There are two minimal winning coalitions [4,2,3:6] -A is in both -B is in only one -C is in only one A = (½)*(½ + ½) = ½; B = C = (½) * (½)=1/4

B C Deegan-Packel Power Distribution

Quota Deegan-Packel Another example There are three minimal winning coalitions [4,2,3,1:4] -A is in one (of size 1) -B is in only one (of size two) -C is in two(of size two) -D is in one (of size two) A = (1/3)*(1)= 1/3; B =(1/3) * (½)=1/6 C=(1/3)* (½ + ½) = 1/3 D = (1/3) *(1/2) = 1/6 CB A CD

Manipulation of WVGs Annexation Merging

Susceptibility of Power Index to Manipulation Let Φ i (G) be the power of i in G. If there exist an altered game G’ such that Φ i (G’)> Φ i (G). Factor of Increment Φ i (G’) /Φ i (G) Domination of Manipulability Let Φ and θ be two power indices. If the increment of Φ w.r.t G and G’ is greater than θ. Then Φ dominates θ, i.e., more susceptible. Important Terms in the Paper

An Example Let G =[5,8,3,3,4,2,4;18] be a WVG of seven agents with agent 1 an annexer. The Deegan-Packel index of the agent 1 in G is Annexation implies that the agent combines with another agent who relinquishes its claim on the power. Suppose the annexer annexes another agent with weight 4. We have a new game G’=[9,8,3,3,2,4;18]. The new Deegan-Packel index of this agent is > , and the factor of increment is 1.51.

Motivation for using Power Indices WVGs have many applications: Economics, Political Science, Neuro Science, Distributed Systems, Multi agent Systems. It is important that games adopt power index which motivates truth telling in order to eliminate the appeal of participating in manipulations. When truth telling is dominant, it provides some assurance of fairness in the games – to the degree that the original power index is fair.

State of the Art WVGs Manipulations False Name Manipulations break into pieces Annexations and Merging Bachrach and Elkind 2008 Azeez and Paterson 2009 Lasisi and Allan 2010 Machover and Felsenthal 2002 Azeez and Paterson 2009

About the Paper We consider agents engaging in Annexation and Merging in WVGs. We evaluate agents’ power using Shapley-Shubik, Banzhaf, and Deegan-Packel Indices. A WVG in which there is a single winning coalition and every agent is critical to the coalition is a Unanimity WVG. All voters must be present to form a winning coalition. We consider Unanimity and Non Unanimity WVGs.

Original Contributions-Unanimity WVGs  Annexation increases the power of other agents (that are not annexed) by the same factor of increment as the annexer. In other words, if any agent annexes, all benefit as they are all equally critical and there are fewer agents to split the power. The annexer also incurs annexation costs, thus reducing its benefits.  All of the indices are affected by annexation. However, the manipulability of any one type of power index does not dominate the manipulability of other types of indices.  Given that there are n agents in the original game, the upper bound on the extent to which a strategic agent may gain is at most n times the power of the agent in the original game.

Experiment We have 15 agents. We annex anywhere from 1-10 other agents (termed the bloc size). The block size is randomly generated as are the agents which are annexed. The weight of the annexer is the sum of its original weight and the weight of the annexed agents.

Original Contributions– Non Unanimity WVGs Annexation Figure: Susceptibility to Manipulation via Annexation

Original Contributions– Non Unanimity WVGs Merging Figure: Susceptibility to Manipulation via Merging

Interpretation Only Shapley-Shubik appears to be susceptible to manipulation via merging. There appears to be no correlation between block size and factor of increment, so would be manipulators would need to use trial and error. The relative susceptibility between the indices is clear. The highest average factor of increment is less than 1.

Original Contribution– Non Unanimity WVGs Merging Figure: Percentage of Advantageous and Disadvantageous Games for Manipulation via Merging

Conclusions The three indices show various degrees of susceptibility to manipulations via annexation and merging with Shapley-Shubik being the most susceptible for both annexation and Merging For unanimity WVGs of n agents, the upper bound on the extent to which a strategic agent may gain is n times its power in the original game. For non unanimity WVGs, the games are less vulnerable to manipulation via merging, while they are extremely vulnerable to manipulation via annexation. Finally, in relation to Lasisi and Allan 2010, the situation where splitting by a strategy is disadvantageous corresponds to situation where it is advantageous for several strategic agents to merge.

Future Work We have assumed in the paper that for the case of merging in non Unanimity WVGs, agents can easily distribute their gains in a fair and stable manner. We plan to investigate the assumption using Game- theoretic approach if there exists such stable and fair ways of distributing the gains.