Piotr Faliszewski (AGH University of
Science and Technology)
There is a number of ways of generating synthetic preference
data. For example, one might choose each possible vote uniformly at
random, or use some urn model (with a given parameter of contagion) or
(a mixture of) the Mallows model(s), with a given parameter of central
coherence, or some Euclidean model, parametrized by the distribution
of the ideal points of candidates and voters. However, if one is to
evaluate an election-related algorithm, it is not clear which of these
models to use, and for what parameters. In this talk I will present a
principled approach to answering this question and I will show very
colorful pictures showing that I am right.
Back to the
research colloquium site.