While there is evidence that user-adaptive support can greatly enhance the effectiveness of educational systems, designing such support for exploratory learning environments (e.g., simulations) is still challenging due to the open-ended nature of their interaction. In particular, there is little a priori knowledge of which student's behaviors can be detrimental to learning in such environments. To address this problem, we focus on a data-driven user-modeling framework that uses logged interaction data to learn which behavioral or activity patterns should trigger help during interaction with a specific learning environment. This framework has been successfully used to provide adaptive support in interactive learning simulations. Here we present a novel application of this framework we are working on, namely to Massive Open Online Courses (MOOCs), a form of exploratory environment that could greatly benefit from adaptive support due to the large diversity of their users, but typically lack of such adaptation. We describe an experiment aimed at investigating the value of our framework to identify student's behaviors that can justify adapting to, and report some preliminary results.