Representation Learning for Latent Risk Dynamics in Online Gambling

by Mr Sam Andersson (Karolinska Institutet)

Wednesday 13 May 2026, 15:00 → 16:00 Europe/Stockholm

AlbaNova A5:1041 - CoPS grupprum (AlbaNova Main Building)

Many digital systems generate behavioural time series that are heavy-tailed, bursty, and only selectively labelled, making it difficult to define a stable risk state from routine data. In this talk, I use online gambling as a case study and present a modelling framework that combines tail-focused exceedance features, representation learning, and dynamic Bayesian state modelling to derive an auditable operational proxy definition of gambling-related risk. The aim is not clinical diagnosis, but a practical and transparent way to represent time-varying behavioural regimes and evaluate early-warning policies under realistic review-capacity constraints.

Zoom link: https://stockholmuniversity.zoom.us/j/69445559626

About the speaker: Sam Andersson is a doctoral researcher at Karolinska Institutet focusing on computational statistics, probabilistic modelling and machine learning for longitudinal behaviour data.