Chair of Artificial Intelligence and Machine Learning

Breadcrumb Navigation


Single-Peaked Preferences for Modeling Political Orientation (Ba/Ma)

Topic for a bachelor/master's thesis

Short Description:

The notion of single-peaked preferences refers to a rather simple (and restricted) type of utility function. Roughly speaking, preferences are single-peaked if a set of choice alternatives can be sorted in a single dimension “from left to right”, so that one of these alternatives is assigned the highest degree of utility, and the utility for the other alternatives decreases with increasing distance from that alternative in both directions [1].

The goal of the thesis is to apply models of that kind to a data set comprising voters’ opinions in the form of pairwise comparisons between political parties (in Germany). The main question behind is whether political orientation is sufficiently “one-dimensional”, or maybe more complex, so that additional dimensions are needed to characterize the orientation of voters and provide an adequate picture of the political landscape. To this end, algorithms must be implemented for fitting single-peaked preference functions to the data, as well as goodnessof-fit tests to determine the quality of these preference models.

The thesis is planned as a cooperation with the Chair of Empirical Political Research and Policy Analysis (Prof. P. Thurner).


Good background in statistics, machine learning and data analysis; algorithm design; programming skills.


Prof. Eyke Hüllermeier


  • [1] [1] V. Conitzer. Eliciting Single-Peaked Preferences Using Comparison Queries. Journal of Artificial Intelligence Research, 35, 2009.