Lecture 12: Topics in Voting Theory

Similar documents
Introduction to Theory of Voting. Chapter 2 of Computational Social Choice by William Zwicker

Computational Social Choice: Spring 2007

Exercises For DATA AND DECISIONS. Part I Voting

(67686) Mathematical Foundations of AI June 18, Lecture 6

Voting Systems That Combine Approval and Preference

Computational Social Choice: Spring 2017

Safe Votes, Sincere Votes, and Strategizing

Voting rules: (Dixit and Skeath, ch 14) Recall parkland provision decision:

Chapter 9: Social Choice: The Impossible Dream Lesson Plan

Introduction to the Theory of Voting

The Manipulability of Voting Systems. Check off these skills when you feel that you have mastered them.

Chapter 10. The Manipulability of Voting Systems. For All Practical Purposes: Effective Teaching. Chapter Briefing

Voting Criteria: Majority Criterion Condorcet Criterion Monotonicity Criterion Independence of Irrelevant Alternatives Criterion

Arrow s Impossibility Theorem

Social Choice & Mechanism Design

Rationality of Voting and Voting Systems: Lecture II

1.6 Arrow s Impossibility Theorem

Topics on the Border of Economics and Computation December 18, Lecture 8

Fairness Criteria. Majority Criterion: If a candidate receives a majority of the first place votes, that candidate should win the election.

Critical Strategies Under Approval Voting: Who Gets Ruled In And Ruled Out

Social welfare functions

Social Choice. CSC304 Lecture 21 November 28, Allan Borodin Adapted from Craig Boutilier s slides

Approaches to Voting Systems

CS 886: Multiagent Systems. Fall 2016 Kate Larson

Voting System: elections

12.2 Defects in Voting Methods

Voter Response to Iterated Poll Information

Chapter 1 Practice Test Questions

Voter Sovereignty and Election Outcomes

Fairness Criteria. Review: Election Methods

CSC304 Lecture 14. Begin Computational Social Choice: Voting 1: Introduction, Axioms, Rules. CSC304 - Nisarg Shah 1

How should we count the votes?

Algorithms, Games, and Networks February 7, Lecture 8

Public Choice. Slide 1

Voting Protocols. Introduction. Social choice: preference aggregation Our settings. Voting protocols are examples of social choice mechanisms

CSC304 Lecture 16. Voting 3: Axiomatic, Statistical, and Utilitarian Approaches to Voting. CSC304 - Nisarg Shah 1

Democratic Rules in Context

MATH 1340 Mathematics & Politics

Desirable properties of social choice procedures. We now outline a number of properties that are desirable for these social choice procedures:

Economics 470 Some Notes on Simple Alternatives to Majority Rule

Voting and preference aggregation

Recall: Properties of ranking rules. Recall: Properties of ranking rules. Kenneth Arrow. Recall: Properties of ranking rules. Strategically vulnerable

Intro to Contemporary Math

Measuring Fairness. Paul Koester () MA 111, Voting Theory September 7, / 25

Social Choice Theory. Denis Bouyssou CNRS LAMSADE

Many Social Choice Rules

Lecture 16: Voting systems

Voting and preference aggregation

Rock the Vote or Vote The Rock

Mathematics and Social Choice Theory. Topic 4 Voting methods with more than 2 alternatives. 4.1 Social choice procedures

Notes for Session 7 Basic Voting Theory and Arrow s Theorem

Varieties of failure of monotonicity and participation under five voting methods

answers to some of the sample exercises : Public Choice

Chapter 9: Social Choice: The Impossible Dream

Head-to-Head Winner. To decide if a Head-to-Head winner exists: Every candidate is matched on a one-on-one basis with every other candidate.

Mathematical Thinking. Chapter 9 Voting Systems

VOTING TO ELECT A SINGLE CANDIDATE

Australian AI 2015 Tutorial Program Computational Social Choice

SOCIAL CHOICES (Voting Methods) THE PROBLEM. Social Choice and Voting. Terminologies

Electing the President. Chapter 12 Mathematical Modeling

Experimental Evidence on Condorcet-Eciency and Strategic Voting: Plurality vs Approval Voting

MATH4999 Capstone Projects in Mathematics and Economics Topic 3 Voting methods and social choice theory

The search for a perfect voting system. MATH 105: Contemporary Mathematics. University of Louisville. October 31, 2017

NP-Hard Manipulations of Voting Schemes

The Impossibilities of Voting

Voting Systems for Social Choice

Strategic voting. with thanks to:

Strategic Voting and Strategic Candidacy

Trying to please everyone. Ulle Endriss Institute for Logic, Language and Computation University of Amsterdam

The Mathematics of Voting. The Mathematics of Voting

Voting and Markov Processes

Chapter 2 Descriptions of the Voting Methods to Be Analyzed

Kybernetika. Robert Bystrický Different approaches to weighted voting systems based on preferential positions

Chapter 9: Social Choice: The Impossible Dream Lesson Plan

Write all responses on separate paper. Use complete sentences, charts and diagrams, as appropriate.

Make the Math Club Great Again! The Mathematics of Democratic Voting

Voting: Issues, Problems, and Systems, Continued

Intro Prefs & Voting Electoral comp. Voter Turnout Agency GIP SIP Rent seeking Partisans. Political Economics. Dr. Marc Gronwald Dr.

Social choice theory

Introduction: The Mathematics of Voting

Voting and Complexity

Elections with Only 2 Alternatives

Introduction to Social Choice

Social Choice: The Impossible Dream. Check off these skills when you feel that you have mastered them.

Name Date I. Consider the preference schedule in an election with 5 candidates.

Math for Liberal Arts MAT 110: Chapter 12 Notes

Constructing voting paradoxes with logic and symmetry

2-Candidate Voting Method: Majority Rule

Voting Lecture 3: 2-Candidate Voting Spring Morgan Schreffler Office: POT Teaching.

A Minimax Procedure for Electing Committees

Voting Definitions and Theorems Spring Dr. Martin Montgomery Office: POT 761

9.3 Other Voting Systems for Three or More Candidates

Rationality & Social Choice. Dougherty, POLS 8000

A Minimax Procedure for Negotiating Multilateral Treaties

The Math of Rational Choice - Math 100 Spring 2015

A Framework for the Quantitative Evaluation of Voting Rules

The Mathematics of Voting

Main idea: Voting systems matter.

Problems with Group Decision Making

HANDBOOK OF SOCIAL CHOICE AND VOTING Jac C. Heckelman and Nicholas R. Miller, editors.

Transcription:

Lecture 12: Topics in Voting Theory Eric Pacuit ILLC, University of Amsterdam staff.science.uva.nl/ epacuit epacuit@science.uva.nl Lecture Date: May 11, 2006 Caput Logic, Language and Information: Social Software staff.science.uva.nl/ epacuit/caputlli.html

Introduction Introduction to the mathematics of voting procedures Two new voting procedures Can polls change an election outcome? Moulin. Axioms of Cooperative Decision Making (Chapter 9). Cambridge University Press (1988). Brams and Fishburn. Voting Procedures. Handbook of Social Choice and Welfare (2002). For a geometric perspective see D. Saari. Basic Geometry of Voting. Springer (1995).

Voting Problem Given a (nite!) set X of candidates and a (nite!) set A of voters each of whom have a preference over X Devise a method F which aggregates the individual preferences to produce a collective decision (typically a subset of X)

Voting Procedures Type of vote, or ballot, that is recognized as admissible by the procedure: let B(X) be the set of admissible ballots for a set X of candidates A method to count a vector of ballots (one ballot for each voter) and select a winner (or winners)

Voting Procedures Type of vote, or ballot, that is recognized as admissible by the procedure: let B(X) be the set of admissible ballots for a set X of candidates A method to count a vector of ballots (one ballot for each voter) and select a winner (or winners) Formally, A voting procedure for a set A of agents (with A = n) and a set X of candidates is a pair (B(X), Ag) B(X) is a set of ballots; and Ag : B(X) n 2 X (typically we are interested in the case where Ag( b) = 1).

Examples Plurality (Simple Majority) B(X) = X Given b X n and x X, let #x( b) = {i bi=x} 1 Ag( b) = {x #x( b) is maximal}

Examples Plurality (Simple Majority) B(X) = X Given b X n and x X, let #x( b) = {i bi=x} 1 Ag( b) = {x #x( b) is maximal} Approval Voting B(X) = 2 X Ag( b) = {x #x( b) is maximal}

Comparing Voting Procedures Does a procedure truly reect the will of the population? Some Properties: Pareto Optimality: If a candidate x is unanimously preferred to candidate b, then b should not be elected Anonymity: The name of the voters does not matter (if two voters change votes, then the outcome is unaected) Neutrality: The name of the candidates does not matter (if two candidate are exchanged in every ranking of the candidates, then the outcome changes accordingly) Monotonicity: Moving up in the rankings is always better

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Plurality Winner: a with 8 votes

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Plurality Winner: a with 8 votes However, a strong majority 13 rank a last!

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Condorcet Winner: c beats every other candidate in a pairwise election.

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Condorcet Winner: c beats every other candidate in a pairwise election. c vs. b

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Condorcet Winner: c beats every other candidate in a pairwise election. c vs. d

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Condorcet Winner: a candidate a such that for all b a, more voters prefer a over b than b over a Condorcet Consistent Voting Rule: A voting rule that always elects the Condorcet winner (if one exists)

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Borda: Take into acount the entire ordering.

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Borda: Take into acount the entire ordering. b vs. c

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Borda: Take into acount the entire ordering. b vs. c 7 rank b rst 6 rank c rst 8 rank a rst

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Borda: Take into acount the entire ordering. b vs. c 16 rank b rst or second 11 rank c rst or second

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Borda: Take into acount the entire ordering. b vs. c all voters rank b rst, second or third all voters rank c rst, second or third

Reecting the will of the population # voters 3 5 7 6 a a b c b c d b c b c d d d a a Borda: Take into acount the entire ordering. b vs. c Thus, b better reects the preferences of the population.

Borda Count # voters 3 5 7 6 a a b c b c d b c b c d d d a a

Borda Count # voters 3 5 7 6 3 a a b c 2 b c d b 1 c b c d 0 d d a a Borda: Take into acount the entire ordering. BC(a) = 3 3 + 3 5 + 0 7 + 0 6 = 24 BC(b) = 2 3 + 1 5 + 3 7 + 2 6 = 44 BC(c) = 1 3 + 2 5 + 1 7 + 3 6 = 29 BC(d) = 0 3 + 0 5 + 2 7 + 1 6 = 20

Borda Count # voters 3 5 7 6 3 a a b c 2 b c d b 1 c b c d 0 d d a a Borda: Take into acount the entire ordering. BC(a) = 3 3 + 3 5 + 0 7 + 0 6 = 24 BC(b) = 2 3 + 1 5 + 3 7 + 2 6 = 44 BC(c) = 1 3 + 2 5 + 1 7 + 3 6 = 29 BC(d) = 0 3 + 0 5 + 2 7 + 1 6 = 20

Borda Count # voters 3 5 7 6 3 a a b c 2 b c d b 1 c b c d 0 d d a a Borda: Take into acount the entire ordering. BC(a) = 3 3 + 3 5 + 0 7 + 0 6 = 24 BC(b) = 2 3 + 1 5 + 3 7 + 2 6 = 44 BC(c) = 1 3 + 2 5 + 1 7 + 3 6 = 29 BC(d) = 0 3 + 0 5 + 2 7 + 1 6 = 20

Scoring Rule Fix a nondecreasing sequence of real numbers s1 s1 sm 1 with s0 < sm 1 Voters rank the candidates, thus giving s0 points to the one ranked last, s1 to the one ranked next to last, and so on. A candidate with the maximal totla score is elected. Theorem (Fishburn) There are proels where the Condorcet winner is never elected by any scoring method.

Failure of Monotonicity Plurality with runo: In the rst round each voter casts a vote for one candidate. If a conadidate wins a strict majority of voites, he is elected. Otherwise, a runo by majority voting is called between the two candidates that received the most votes in the rst round.

Failure of Monotonicity Plurality with runo: In the rst round each voter casts a vote for one candidate. If a conadidate wins a strict majority of voites, he is elected. Otherwise, a runo by majority voting is called between the two candidates that received the most votes in the rst round. # voters 6 5 4 2 a c b b b a c a c b a c # voters 6 5 4 2 a c b a b a c b c b a c

Failure of Monotonicity Plurality with runo: In the rst round each voter casts a vote for one candidate. If a conadidate wins a strict majority of voites, he is elected. Otherwise, a runo by majority voting is called between the two candidates that received the most votes in the rst round. # voters 6 5 4 2 a c b b b a c a c b a c # voters 6 5 4 2 a c b a b a c b c b a c The proles are monotonic.

Failure of Monotonicity Plurality with runo: In the rst round each voter casts a vote for one candidate. If a conadidate wins a strict majority of voites, he is elected. Otherwise, a runo by majority voting is called between the two candidates that received the most votes in the rst round. # voters 6 5 4 2 a c b b b a c a c b a c # voters 6 5 4 2 a c b a b a c b c b a c a wins with the rst prole, but c wins with the second.

Failure of Monotonicity Plurality with runo: In the rst round each voter casts a vote for one candidate. If a candidate wins a strict majority of votes, he is elected. Otherwise, a runo by majority voting is called between the two candidates that received the most votes in the rst round. # voters 6 5 4 2 a x b b b a x a x b a x # voters 6 5 4 2 a c x a x a c x c x a c a wins with the rst prole, but c wins with the second.

No-show Paradox Totals Rankings H over W W over H 417 B H W 417 0 82 B W H 0 82 143 H B W 143 0 357 H W B 357 0 285 W B H 0 285 324 W H B 0 324 1608 917 691 Fishburn and Brams. Paradoxes of Preferential Voting. Mathematics Magazine (1983).

No-show Paradox Totals Rankings H over W W over H 417 B H W 417 0 82 B W H 0 82 143 H B W 143 0 357 H W B 357 0 285 W B H 0 285 324 W H B 0 324 1608 917 691 B: 417 + 82 = 499 H: 143 + 357 = 500 W: 285 + 324 = 609

H Wins No-show Paradox Totals Rankings H over W W over H 417 X H W 417 0 82 X W H 0 82 143 H X W 143 0 357 H W X 357 0 285 W X H 0 285 324 W H X 0 324 1608 917 691

No-show Paradox Totals Rankings H over W W over H 419 B H W 417 0 82 B W H 0 82 143 H B W 143 0 357 H W B 357 0 285 W B H 0 285 324 W H B 0 324 1610 917 691 Suppose two more people show up with the ranking B H W

No-show Paradox Totals Rankings H over W W over H 419 B H W 417 0 82 B W H 0 82 143 H B W 143 0 357 H W B 357 0 285 W B H 0 285 324 W H B 0 324 1610 917 691 B: 419 + 82 = 501 H: 143 + 357 = 500 W: 285 + 324 = 609

No-show Paradox Totals Rankings B over W W over B 419 B X W 419 0 82 B W X 82 82 143 X B W 143 0 357 X W B 0 357 285 W B X 0 285 324 W X B 0 324 1610 644 966 B: 419 + 82 = 501 H: 143 + 357 = 500 W: 285 + 324 = 609

W Wins! No-show Paradox Totals Rankings B over W W over B 419 B X W 419 0 82 B W X 82 82 143 X B W 143 0 357 X W B 0 357 285 W B X 0 285 324 W X B 0 324 1610 644 966

Multiple Districts Totals Rankings East West 417 B H W 160 257 82 B W H 0 82 143 H B W 143 0 357 H W B 0 357 285 W B H 0 285 324 W H B 285 39 1608 588 1020

Multiple Districts Totals Rankings East West 417 B H W 160 257 82 B W H 0 82 143 H B W 143 0 357 H W B 0 357 285 W B H 0 285 324 W H B 285 39 1608 588 1020 B would win both districts!

Young's Theorem Reinforcement: If two disjoint groups of voters N1 and N2 face the saem set of candidates and Ni selects Bi. If B1 B2, then N1 N2 should select B1 B2. Continuity Suppose N1 elects candidate a and a disjoint group N2 elects b a. Then there is a m such that (nn1) N2 chooses a. Theorem (Young) A voting correspondence is a scoring method i it satises anonymity, neutrality, reinforcement and continuity. Young. Social Choice Scoring Functions. SIAM Journal of Applied Mathematics (1975).

Approval Voting Theorem (Fishburn) A voting correspondence is approval voting i it satises anonymity, neutrality, reinforcement and If a prole consists of exactly two ballots (sets of candidates) A and B with A B =, then the procedure selects A B. Fishburn. Axioms for Approval Voting: Direct Proof. Journal of Economic Theory (1978).

Strategic Voting Theorem (Gibbard (1973) and Satterthwaite (1975)) If there are at lest three candidates, a voting rule is strategyproof i it is dictatorial. We will not discuss these issues today. See A. Taylor. Social Choice and the Mathematics of Strategizing. 2005.

Digression: Fall-Back Bargaining Voter Type d=0 d=1 d=2 d=3 Sum Max ( 1, 1, 1 ) 0 9 12 13 18 3 ( 1, 1, 0 ) 3 5 12 13 19 3 ( 1, 0, 1 ) 3 5 12 13 19 3 ( 0, 1, 1 ) 3 5 12 13 19 3 ( 1, 0, 0 ) 1 8 10 13 20 3 ( 0, 1, 0 ) 1 8 10 13 20 3 ( 0, 0, 1 ) 1 8 10 13 20 3 ( 0, 0, 0 ) 1 4 13 13 21 2 Total 13 52 91 104 FB (Minimax): ( 0, 0, 0 ), MV (Minisum): ( 1, 1, 1 )

Combining Approval and Preference Under Approval Voting (AV), voters are asked which candidates the voter approves

Combining Approval and Preference Under Approval Voting (AV), voters are asked which candidates the voter approves Under preference voting procedures (such as BC), candidates are asked to (linearly) rank the candidates.

Combining Approval and Preference Under Approval Voting (AV), voters are asked which candidates the voter approves Under preference voting procedures (such as BC), candidates are asked to (linearly) rank the candidates. The two pieces of information are related, but not derivable from each other

Combining Approval and Preference Under Approval Voting (AV), voters are asked which candidates the voter approves Under preference voting procedures (such as BC), candidates are asked to (linearly) rank the candidates. The two pieces of information are related, but not derivable from each other What about asking for both pieces of information?

Combining Approval and Preference Under Approval Voting (AV), voters are asked which candidates the voter approves Under preference voting procedures (such as BC), candidates are asked to (linearly) rank the candidates. The two pieces of information are related, but not derivable from each other What about asking for both pieces of information? Brams and Sanver. Voting Systems that Combine Approval and Preference. available at the author's website.

Assumptions Assume each voter has a (linear) preference over the candidates. Each voter is asked to rank the candidates from most preferred to least preferred (ties are not allowed). Voters are then asked to specify which candidates are acceptable. Consistency Assumption Given two candidates a and b, if a is approved and b is disapproved then a is ranked higher than b. For example, we denote this approval ranking for a set {a, b, c, d} of candidates as follows a d c b

Preference Approval Voting (PAV) 1. If no candidate, or exactly one candidate, receives a majority of approval votes, the PAV winner is the AV winner.

Preference Approval Voting (PAV) 1. If no candidate, or exactly one candidate, receives a majority of approval votes, the PAV winner is the AV winner. 2. If two or more candidates receive a majority of approval votes, then (a) If one of these candidates is preferred by a majority to every other majority approved candidate, then he or she is the PAV winner.

Preference Approval Voting (PAV) 1. If no candidate, or exactly one candidate, receives a majority of approval votes, the PAV winner is the AV winner. 2. If two or more candidates receive a majority of approval votes, then (a) If one of these candidates is preferred by a majority to every other majority approved candidate, then he or she is the PAV winner. (b) If there is not one majority-preferred candidate because of a cycle among the majority-approved candidates, then the AV winner among them is the PAV winner.

Rule 1 PAV vs. Condorcet I. 1 voter: a b c II. 1 voter: b a c III. 1 voter: c a b

Rule 1 b is the AV winner. PAV vs. Condorcet I. 1 voter: a b c II. 1 voter: b a c III. 1 voter: c a b

PAV vs. Condorcet Rule 1 I. 1 voter: a b c II. 1 voter: b a c III. 1 voter: c a b b is the AV winner. b is also the PAV winner.

PAV vs. Condorcet Rule 1 I. 1 voter: a b c II. 1 voter: b a c III. 1 voter: c a b b is the AV winner. b is also the PAV winner. a is the Condorcet winner.

Rule 2(a) PAV vs. Condorcet I. 1 voter: a b c d II. 1 voter: b c a d III. 1 voter: d a c b

Rule 2(a) b is the PAV winner. PAV vs. Condorcet I. 1 voter: a b c d II. 1 voter: b c a d III. 1 voter: d a c b

PAV vs. Condorcet Rule 2(a) I. 1 voter: a b c d II. 1 voter: b c a d III. 1 voter: d a c b b is the PAV winner. a is the Condorcet winner.

Rule 2(b) PAV vs. Condorcet I. 1 voter: d a b c e II. 1 voter: d b c a e III. 1 voter: e d c a b IV. 1 voter: a b c d e V. 1 voter: c b a d e

PAV vs. Condorcet Rule 2(b) I. 1 voter: d a b c e II. 1 voter: d b c a e III. 1 voter: e d c a b IV. 1 voter: a b c d e V. 1 voter: c b a d e a (3 votes), b (3 votes), and c (4 votes) are all majority approved.

PAV vs. Condorcet Rule 2(b) I. 1 voter: d a b c e II. 1 voter: d b c a e III. 1 voter: e d c a b IV. 1 voter: a b c d e V. 1 voter: c b a d e a (3 votes), b (3 votes), and c (4 votes) are all majority approved. c is the PAV winner

Rule 2(b) PAV vs. Condorcet I. 1 voter: d a b c e II. 1 voter: d b c a e III. 1 voter: e d c a b IV. 1 voter: a b c d e V. 1 voter: c b a d e a (3 votes), b (3 votes), and c (4 votes) are all majority approved. c is the PAV winner. d is the Condorcet winner.

Example I. 3 voters: a b c d II. 3 voters: d a c b III. 2 voters: b d c a

Example I. 3 voters: a b c d II. 3 voters: d a c b III. 2 voters: b d c a c is approved by all 8 voters.

Example I. 3 voters: a b c d II. 3 voters: d a c b III. 2 voters: b d c a c is approved by all 8 voters. There is a top cycle a > b > d > a which are all preferred by majorities to c (the AV winner). a is the PAV winner

a is the PAV winner. c is the AV winner. d is the STV winner. Example I. 3 voters: a b c d II. 3 voters: d a c b III. 2 voters: b d c a

Example I. 2 voters: a c b d II. 2 voters: a c d b III. 3 voters: b c d a

Example I. 2 voters: a c b d II. 2 voters: a c d b III. 3 voters: b c d a c is approve by all 7 voters. a is the least approved candidate. a is the PAV winner. BC(a) = 12 BC(c) = 14

Fallback Voting (FV) 1. Voters indicate all candidates of whom they approve, who may range from no candidate to all candidate. Voters rank only those candidates whom they approve. 2. The highest-ranked candidate of all voters is considered. If a majority agrre on the highest-ranked candidate, this candidate is the FV winner (level 1). 3. If there is no level 1 winner, the next-highest ranked candidate of all voters in considered. If a majority of voters agree on one candidate as either their highest or their next-highest ranked candidate, this candidiate is the FV winner (level 2). If more than one receive majority approval, then the candidate with the largest majority is the FV winner.

4. If no level 2 winner, the voters descend one level at a time to lower ranks of approved candidates stopping when one or more candidates receives majority approval. If more than one receives majority approval then the candidate with the largest majority is the FV winner. If the descent reaches the bottom and no candidate has won, then the candidate with the most approval is the FV winner.

Example I. 4 voters: a b c d II. 3 voters: b c a d III. 2 voters: d a c b

b is the FV winner. c is the AV winner. a is the PAV winner. Example I. 4 voters: a b c d II. 3 voters: b c a d III. 2 voters: d a c b

PAV and FV Neither PAV nor FV may elect the Condorcet winner. Both PAV and FV are monotonic (approval-monotonic and rank-monotonic) Truth-telling strategies of voters under PAV and FV may not be in equilibrium

Polls in elections How can polls change the outcome of an election?

Example I The following example is due to Brams and Fishburn P A = o1 > o3 > o2 P B = o2 > o3 > o1 P C = o3 > o1 > o2 Size Group I II 4 A o1 o1 3 B o2 o2 2 C o3 o1 If the current winner is o, then agent i will switch its vote to some candidate o provided 1. o is one of the top two candidates as indicated by a poll 2. o is preferred to the other top candidate

Example I The following example is due to [BF] P A = o1 > o3 > o2 P B = o2 > o3 > o1 P C = o3 > o1 > o2 Size Group I II 4 A o1 o1 3 B o2 o2 2 C o3 o1 If the current winner is o, then agent i will switch its vote to some candidate o provided 1. o is one of the top two candidates as indicated by a poll 2. o is preferred to the other top candidate

Example II P A = (o1, o4, o2, o3) P B = (o2, o1, o3, o4) P C = (o3, o2, o4, o1) P D = (o4, o1, o2, o3) P E = (o3, o1, o2, o4) Size Group I II III IV 40 A o1 o1 o4 o1 30 B o2 o2 o2 o2 15 C o3 o2 o2 o2 8 D o4 o4 o1 o4 7 E o3 o3 o1 o1 If the current winner is o, then agent i will switch its vote to some candidate o provided 1. i prefers o to o, and 2. the current total for o plus agent i's votes for o is greater than the current total for o.

Example II P A = (o1, o4, o2, o3) P B = (o2, o1, o3, o4) P C = (o3, o2, o4, o1) P D = (o4, o1, o2, o3) P E = (o3, o1, o2, o4) Size Group I II III IV 40 A o1 o1 o4 o1 30 B o2 o2 o2 o2 15 C o3 o2 o2 o2 8 D o4 o4 o1 o4 7 E o3 o3 o1 o1 If the current winner is o, then agent i will switch its vote to some candidate o provided 1. i prefers o to o, and 2. the current total for o plus agent i's votes for o is greater than the current total for o.

Example II P A = (o1, o4, o2, o3) P B = (o2, o1, o3, o4) P C = (o3, o2, o4, o1) P D = (o4, o1, o2, o3) P E = (o3, o1, o2, o4) Size Group I II III IV 40 A o1 o1 o4 o1 30 B o2 o2 o2 o2 15 C o3 o2 o2 o2 8 D o4 o4 o1 o4 7 E o3 o3 o1 o1 If the current winner is o, then agent i will switch its vote to some candidate o provided 1. i prefers o to o, and 2. the current total for o plus agent i's votes for o is greater than the current total for o.

Example II P A = (o1, o4, o2, o3) P B = (o2, o1, o3, o4) P C = (o3, o2, o4, o1) P D = (o4, o1, o2, o3) P E = (o3, o1, o2, o4) Size Group I II III IV 40 A o1 o1 o4 o1 30 B o2 o2 o2 o2 15 C o3 o2 o2 o2 8 D o4 o4 o1 o4 7 E o3 o3 o1 o1 If the current winner is o, then agent i will switch its vote to some candidate o provided 1. i prefers o to o, and 2. the current total for o plus agent i's votes for o is greater than the current total for o.

Towards a Formal Model A formal model is proposed in which the notion of a protocol is made formal. Agents change their current vote based on two pieces of information: 1. The poll information 2. The number of agents in i's group Formal details can be found in Chapter 5 of my thesis.

Conclusions Towards a theory of correctness of social procedures.

Conclusions Towards a theory of correctness of social procedures. Using logic as a tool to analyze game theoretic and social choice problems.

Conclusions Towards a theory of correctness of social procedures. Using logic as a tool to analyze game theoretic and social choice problems. There are a wealth of social procedures described in the game theory literature. Attempts to formalize these procedures create a wide range of interesting problems for logicians. It will be very interesting to see whether a logical analysis can help create new procedures or rene old ones.

Conclusions Logical Omniscience Problem. Deliberation, to the extent that it is thought of as a rational process of guring out what one should do given one's priorities and expectations is an activity that is unnecessary for the deductively omniscient. Robert Stalnaker

Conclusions Logical Omniscience Problem. Deliberation, to the extent that it is thought of as a rational process of guring out what one should do given one's priorities and expectations is an activity that is unnecessary for the deductively omniscient. Robert Stalnaker Empirical Studies

Final paper is due May 29, 2006 Send the paper to epacuit@science.uva.nl.