The root system of a linear voting method

03/07/2017 - 12:15pm
03/07/2017 - 1:10pm
Prasad Senesi (Catholic University of America)

In an election with ballots consisting of full rankings of n candidates, the Borda Count voting method provides an aggregate numerical ranking of the candidates. This method is naturally generalized by replacing the standard weights of the Borda Count by a weight vector in an n-dimensional vector space, yielding the so-called positional voting methods, or by replacing fully-ranked ballots with those in the shape of a composition of n, with multiple positions available for each ’place’. Building upon a vector space of profiles introduced by Donald Saari in the 1990’s, Michael Orrison and colleagues used methods from the representation theory of the symmetric group’s action on compositions to study these positional and other voting methods. In this talk we will use the standard Euclidean inner product on this vector space to show how the neutrality of a positional voting method is combinatorially manifested by the appearance of a type-A root system in this vector space of profiles, and conversely how this root system can be used to construct any neutral positional voting method.

Millikan 2099, Pomona College