The needs of mobile learners change dynamically with the various requirements of their wireless environments highlighting the need for implementing adaptivity in mobile learning. This paper introduces a new approach for adapting content in a mobile environment. This approach suggests adapting the delivery of granules of supplementary learning content to mobile devices using Bayesian belief networks. We describe a framework for producing content granules and controlling the preferences of adaptivity over various probabilistic techniques. The result will be a tool for the deployment and evaluation of adaptivity in a distributed mobile learning environment.