Title: Group Coordination: A History of Paradox and Impossibility
1Group Coordination A History of Paradox and
Impossibility
2Voting Paradox I(Condorcet 1785)
C gt A (2 1)
3Voting Paradox II
Plurality vote A gt B gt C (432)
4How bad can it get?
- Plurality vote A gt B gt C gt D gt E gt gt Z
- Remove Z Y gt X gt W gt V gt U gt gt A or any
other pattern! Saari 95
5Other voting schemes
- Borda count
- Dodgson (Lewis Carroll) winner
- adjacent swap AgtBgtCgtD ? AgtCgtBgtD
- alternative that requires fewest adjacent swaps
to become a Condorcet winner
6Other voting schemes
- Kemeny winner
- d(A,B,gti,gtj) 0 if gti and gtj agree on A,B 1
if one is indiff, the other not 2 if gti and
gtj are opposite - dist(gti,gtj) ?all pairs A,B d(A,B,gti,gtj)
- Winner ordering gt with min ?i dist(gt,gti)
- Dodgson and Kemeny winner are NP-hard!
Bartholdi, Tovey, Trick 89 - Plurality, Borda, Dodgson, Kemeny all depend on
irrelevant alternatives pairwise can lead to
intransitivities.
7How good can it get?
- General case
- gt f(gt1,gt2,...,gtn)
- where gt, gti weak order preference relations
- Q What aggregation function f (e.g., voting
scheme) is independent of irrelevant
alternatives?
A Essentially none!
8Arrows Conditions
- Individual collective rationality gt, gti are
weak orders (transitive) - Universal domain (U)
- Pareto (P) If A gti B for all i, then A gt B
- Indep. of irrelevant alternatives (IIA) gt on
A,B depends only on the gti on A,B - Non-dictatorship (ND) no i s.t. A gti B ? A gt B,
for all A,B
9Arrows Impossibility Theorem
- If persons finite, alternatives gt 2 then
- There is no aggregation function fthat can
simultaneously satisfyU, P, IIA, ND.
10Proof Sketch
- A subgroup G is decisive over A,B if A gti B ,
for all i in G ? A gt B - Field Expansion If G is almost decisive over
A,B, then G is decisive over all pairs. - Group Contraction If any group G is decisive,
then so is some proper subset of G.
11Another Explanation
- IIA ? procedure cannot distinguish btw transitive
intransitive inputs Saari - For example, pairwise vote cannot distinguish
between
A gt B gt C
AgtB, BgtC, CgtA
B gt C gt A
AgtB, BgtC, CgtA
C gt A gt B
AltB, BltC, CltA
12The Impossibility of aParetian Liberal (Sen 1970)
- Liberalism (L) For each i, there is at least one
pair A,B such that A gti B ? A gt B - Minimal Liberalism (L) There are at least two
such free individuals. - There is no aggregation function fthat can
simultaneously satisfyU, P and L. - Does not require IIA.
13Back Doors?
- Fishburn If persons infinite Arrows axioms
are mutually consistent. - But Kirman Sondermann Infinite society
controlled by an arbitrarily small group. An
invisible dictator. - Mihara Determining whether A gt B is uncomputable
14Back Doors?
- Blacks Single-peakedness
- If all voters preferences are single-peaked, then
pairwise (majority) vote satisfiesP, IIA, ND
15Back Doors?
- Cardinal preferences / no interpersonal
comparability ? impossibility remains - Cardinal preferences / interpersonal
comparability ? utilitarianism - u(A) ? ?ui(A)
u1(A) not comparable to u2(C)
16Strategy-proofness(Non-manipulability)
- A voting scheme is manipulable if, in some
situation, it can be advantageous to lie
otherwise it is strategy-proof. - Example Perot gt Clinton gt Bush
- Gibbard and Satterthwaite (independently)
- If of alternatives gt 2,
- Any deterministic, strategy-proof voting scheme
is dictatorial.
17Probabilistic Voting
- Hat of Ballots (HOB) place all ballots in a hat
and choose one top choice at random. - Hat of Alternatives (HOA) Collect ballots.
Choose two alternatives at random. Use any
standard vote to pick one of these two. - HOB HOA are strategy-proof andnon-dictatorial,
but not very appealing. - Gibbard Any strategy-proof voting scheme is a
probability mixture of HOB HOA - (Computing strategy may be intractable B,TT)
18Arrow ? Gibbard-Satterthwaite
- One-to-one correspondence
- Suppose we find a preference aggregation function
f that satisfies U, P, IIA, and ND. - Then the associated vote is strategy-proof
- Suppose we find a strategy-proof vote
- Then an associated f satisfies P, IIA, ND, and U
- Contrapositive another justification for IIA
19Other ImpossibilitiesBelief Aggregation
- Combining probabilities Pr
f(Pr1,Pr2,...,Prn) - Properties / axioms
- Marginalization property (MP)
- Externally Bayesian (EB)
- Proportional Dependence on States (PDS)
- Unanimity (UNAM)
- Independence Preservation Property (IPP)
- Non-dictatorship (ND)
EF
EF
E
EF
EF
EF
EF
/
20Belief Aggregation
- Impossibilities
- IPP, PDS are inconsistent
- MP, EB, UNAM ND are inconsistent
21Other ImpossibilitiesGroup Decision Making
- Setup
- individual probabilities Pri(E), i1,...,n
- individual utilities ui(A?E), i1,...,n
- set of events E
- set of collective actions A
Pr, u
Pr2, u2
Pr3, u3
Pr1, u1
A
E
22Group Decision Making
- Desirable properties / axioms
- (1) Universal domain
- (2) Pr f(Pr1,Pr2,...,Prn) u
g(u1,u2,...,un) - (3) Choice a?A maximizes EU ?EPr(E)u(a,E)
- (4) Pareto Optimal if for all i
EUi(a1)gtEUi(a2), then a2 not chosen - (5) Unanimous beliefs prevail f(Pr,Pr,...,Pr)
Pr - (6) no prob dictator i such that f(Pr1,...,Prn)
Pri - (1)?(6) mutually inconsistent H Z 1979
- does not require IIA
23Other ImpossibilitiesIncentive-compatible trade
- Setup 1 good, 1 buyer w/ value ?a1,b1,seller
w/ value ?a2,b2, nonempty intersect. - Desirable properties / axioms
- (1) incentive compatible
- (2) individually rational
- (3) efficient
- (4) no outside subsidy
- (1)?(4) are inconsistent M S 83
24Other ImpossibilitiesDistributed Computation
- Consensus a fundamental building block
- all processors agree on a value from 0,1
- if all agents choose 0 (1), then output is 0 (1)
- Impossibilities
- unbounded msg delay 1 proc fail by
stopping (common knowledge problem) - no shared mem ?1/3 procs fail
maliciously (Byzantine generals problem)
25Other ImpossibilitiesApportionment
- Setup n congressional seats, pop. of all states
how do we apportion seats to states? - Alabama Paradox
- Desirable properties / axioms
- (1) monotone
- (2) consistent
- (3) satisfying quota
- (1)?(3) are inconsistent B Y 77
26Default Logic
- In default logic, we must sometime choose among
conflicting models - Republicans are by default not pacifists
- Quakers are by default pacifists
- Nixon is both a Republican and a Quaker
- Many conflict resolution strategies
- specificity, chronological, skepticism, credulity
- My default theory M1 gt M2 gt M3
- Your default theory M2 gt M3 gt M1
27Default Logic
- Q Is it possible to construct a universal
default theory, which combines current future
theories? - A No, assuming we want the universal theory to
obey U, P, IIA, ND. - Aside applicability to societies of minds
- Doyle and Wellman 91
28Collaborative Filtering
- Goal predict preferences of one user based on
other users preferences (e.g., movie
recommendations)
29CF and Social Choice
- Usociety f(u1, u2, , un)
- ra f(r1, r2, ... , rn)
- Same functional form
- Similar semantics
- Some of the same constraints on f are desirable,
and have been advocated - Modified limitative theorems are applicableP
H 99
30Ensemble Learning
- censemble f(c1, c2, ... , cn)
- Variants of Arrows thm applies to multiclass
case - Mays axiomatization of majority rule applies to
binary classification case - Common ensemble methods destroy unanimous
independencies - Voting paradoxes can and do occur P,
M-R, G 2000
31Combining Bayesian networks
- Structural unanimity
- Proportional dependence on states
- Pr0(?) ? f(Pr1(?), Pr2(?), , Prn(?))
- Unanimity
- Nondictatorship P W 99
32Combining Bayesian networks
- Structural unanimity
- Proportional dependence on states
- Pr0(?) ? f(Pr1(?), Pr2(?), , Prn(?))
- Unanimity
- Nondictatorship P W 99
33Combining Bayesian networks
- Family aggregation
- Pr0(Epa(E)) fPr1(Epa(E)), , Prn(Epa(E))
- Unanimity
- Nondictatorship P W 99
34Combining Bayesian networks
- Family aggregation
- Pr0(Epa(E)) fPr1(Epa(E)), , Prn(Epa(E))
- Unanimity
- Nondictatorship P W 99
35Conclusion I
- Group coordination is fraught w/ paradox and
impossibilities - voting
- preference aggregation
- belief aggregation
- group decision making
- trading
- distributed computing
- Non-ideal tradeoffs are inevitable
- Standard acceptable solutions seem unlikely
36Conclusion II
- Arrows Theorem initiated social choice theory
remains powerful, compelling - May provide a valuable perspective for computer
scientists interested in multi-agent or
distributed systems
37Simpsons Paradox
- New York
- experiment 54 / 144 (0.375) subjects are cured
- control 12 / 36 (0.333) cured
- California
- experiment 18 / 36 (0.5) cured
- control 66 / 144 (0.458) cured
- Totals
- experiment 70 / 180 cured
- control 78 / 180 cured
38Magic Dice Paradox
gt
gt