A Maximum Likelihood Approach towards Aggregating Partial Orders

In many of the possible applications as well as the theoretical models of computational social choice, the agents' preferences are represented as partial orders. In this paper, we extend the maximum likelihood approach for defining "optimal" voting rules to this setting. We consider distributions in which the pairwise comparisons/incomparabilities between alternatives are drawn i.i.d. We call such models pairwise-independent models and show that they correspond to a class of voting rules that we call pairwise scoring rules. This generalizes rules such as Kemeny and Borda. Moreover, we show that Borda is the only pairwise scoring rule that satisfies neutrality, when the outcome space is the set of all alternatives. We then study which voting rules defined for linear orders can be extended to partial orders via our MLE model. We show that any weakly neutral outcome scoring rule (including any ranking/candidate scoring rule) based on the weighted majority graph can be represented as the MLE of a weakly neutral pairwise-independent model. Therefore, all such rules admit natural extensions to profiles of partial orders. Finally, we propose a specific MLE model πk for generating a set of k winning alternatives, and study the computational complexity of winner determination for the MLE of πk.

[1]  Rolf Niedermeier,et al.  A logic for causal reasoning , 2003, IJCAI 2003.

[2]  Toby Walsh,et al.  mCP Nets: Representing and Reasoning with Preferences of Multiple Agents , 2004, AAAI.

[3]  Bernard De Baets,et al.  A threshold for majority in the context of aggregating partial order relations , 2010, International Conference on Fuzzy Systems.

[4]  H. P. Young,et al.  An axiomatization of Borda's rule , 1974 .

[5]  T. Tideman,et al.  Independence of clones as a criterion for voting rules , 1987 .

[6]  Vincent Conitzer,et al.  Determining Possible and Necessary Winners under Common Voting Rules Given Partial Orders , 2008, AAAI.

[7]  H. Young Social Choice Scoring Functions , 1975 .

[8]  Michel Truchon,et al.  Maximum likelihood approach to vote aggregation with variable probabilities , 2004, Soc. Choice Welf..

[9]  Craig Boutilier,et al.  CP-nets: a tool for represent-ing and reasoning with conditional ceteris paribus state-ments , 2004 .

[10]  Jérôme Lang,et al.  Voting procedures with incomplete preferences , 2005 .

[11]  Vincent Conitzer,et al.  Common Voting Rules as Maximum Likelihood Estimators , 2005, UAI.

[12]  Toby Walsh,et al.  Aggregating Partially Ordered Preferences , 2009, J. Log. Comput..

[13]  Peter C. Fishburn,et al.  Borda's rule, positional voting, and Condorcet's simple majority principle , 1976 .

[14]  Vincent Conitzer,et al.  Generalized scoring rules and the frequency of coalitional manipulability , 2008, EC '08.

[15]  Toby Walsh,et al.  Uncertainty in Preference Elicitation and Aggregation , 2007, AAAI.

[16]  Toby Walsh,et al.  Incompleteness and Incomparability in Preference Aggregation , 2007, IJCAI.

[17]  Nicolas de Condorcet Essai Sur L'Application de L'Analyse a la Probabilite Des Decisions Rendues a la Pluralite Des Voix , 2009 .

[18]  K. Arrow A Difficulty in the Concept of Social Welfare , 1950, Journal of Political Economy.

[19]  Ronen I. Brafman,et al.  CP-nets: A Tool for Representing and Reasoning withConditional Ceteris Paribus Preference Statements , 2011, J. Artif. Intell. Res..

[20]  Michel Truchon Borda and the maximum likelihood approach to vote aggregation , 2008, Math. Soc. Sci..

[21]  Nadja Betzler,et al.  Towards a dichotomy for the Possible Winner problem in elections based on scoring rules , 2009, J. Comput. Syst. Sci..

[22]  Lirong Xia,et al.  Sequential composition of voting rules in multi-issue domains , 2009, Math. Soc. Sci..

[23]  Vincent Conitzer,et al.  Aggregating preferences in multi-issue domains by using maximum likelihood estimators , 2010, AAMAS.