We propose the use of probabilistic programming techniques to model and tackle the malicious user identification problem in a recommendation algorithm. Probabilistic programming provides numerous advantages over other techniques, including but not limited to providing an interpretable latent representation of how malicious users acted under a structured model, as well as allowing for the quantification of damage caused by malicious users, based on counterfactual simulation. We show experiments in malicious user identification and uncertainty quantification using a model of regular and malicious users interacting with a simple recommendation algorithm, and provide a novel simulation-based measure for quantifying the effects of a user or group of users on its dynamics.\

Speaker

Bogdan has been helping us out as a tutor for DATA301 last year and this year. He is a Stanford graduate, worked as a data scientist at Facebook for 7 years, and now runs a Wellington-based data science startup (https://scie.nz/about).