Sie sind hier

User-Centric Analysis of the Liquid Feedback Voting System

Setting:  Liquid Feedback (LQFB) is an online platform where users create petitions and vote in a democratic process that includes delegations (that is, users may delegate their vote to other users).

Data:  We have complete LQFB database dumps from the Pirate Party of Germany with 13348 users, 457362 votes by 6839 distinct users and 9650 delegations (31.01.2013).

Goal: The goal of the thesis will be to analyse the behaviour of users in online voting systems, e.g. to detect the political opinion of individual users by looking at their history of votes.

  • Are there political factions in the Pirate party?
  • Why do people create proposals in the voting system - are they different from users who use the system for voting only?
  • What is the effect of delegations on user behaviour - do delegated users change their behaviour after receiving delegations?
  • Why do people leave the platform?

Methods: For the analysis, statistical methods (e.g. correlation or factor analysis), concepts from network theory (e.g. triangle closing for link prediction) or probabilistic models (such as topic models) can be applied. Detailed prior knowledge on all of these methods is optional!


  • Basic programming skills ( e.g. Java, C, C++ or Perl)
  • Interest in Machine Learning / Data Mining
  • Basic knowledge in statistics

Language: German or English