Abstract
In this paper, we present a prototype framework that combines a custom-made interactive story-like serious game and non-parametric Bayesian modeling (where the generative part of the model is an Agent-based simulation) with a Natural Language Processing (NLP) Deep Learning architecture to label risky decisions. This framework will be used to build an inferential model to categorize real (minor) players upon interaction with an appealing game related to cybercrime and online risk and, more promisingly, to detect risky patterns. The results coming from this research will allow developing science-based policies and educational interventions to protect minors and improve their experience online.
