I am a PhD. candidate at the University of Koblenz and a research assistant at the Computational Social Science department at GESIS. My research interests lie in the fields of Computational Social Science, Network Science and Machine Learning. I currently investigate the effects of network structure on ranking, network inference, and human navigation. I work together with my supervisors Claudia Wagner and Markus Strohmaier, and colleague Fariba Karimi (previously with Philipp Singer and Florian Lemmerich).
The influence of edge formation in ranking: Despite treating individuals equally, ranking algorithms do not guarantee a fair distribution of opportunities among minorities. This motivated me to study when ranking algorithms reinforce inequalities in directed social networks. My approach is to unveil the social mechanisms that can explain inequalities such as skewed distributions and unfair representation of minorities in top ranks using PageRank and Who-To-Follow. My results shed light on the role played by majorities to reduce inequalities without algorithmic intervention in recommender systems. [publications]
The influence of edge formation in classification: Machine learning methods are often utilized as black boxes making it hard to evaluate their accuracy on unseen data. To this end, I have analyzed Bayesian estimation errors and biases in relational classification to provide evaluation benchmarks based on sampling and network structure. Building such benchmarks can help decision makers to enhance their understanding of accuracy and fairness on a particular inference task when no ground-truth is given. [publications]
Edge formation: I have studied how users transition, or create links, from one state to another in different contexts such as: human mobility in the city, and human navigation on the Web. In particular, I have proposed two methods to study navigation: HopRank and Janus. HopRank, a biased random walker, is able to evaluate hypotheses that are based on network structure alone. Its key concept is a model of information foraging on networks based on transition probabilities between k-hop neighborhoods. When metadata is available, Janus, a Bayesian framework, allows to rank hypotheses by relative plausibility. We use it to identify plausible hypotheses in cases where nodes carry additional information. [publications]
I changed my name to Lisette Munz as of date September 2019. However, I will keep my maiden name for academic/publication purposes, and use my new name for legal purposes mainly in Germany.
My maiden name consists of 4 names: 2 first names and 2 last names. This is common in the LatinX community. In my publications you will see a hyphen in between my two last names. Though, it is not official, I do it because otherwise my name gets messed up by certain algorithms.