It is well-known that the Bayesian approach to argumentation (i) has a solid normative foundation and (ii) connects well with empirical data from experiments in the psychology of reasoning and argumentation. The main objective of this research proposal is to demonstrate that it also has the computational resources to allow for large-scale applications in the context of robust argumentation machines. We will adapt some of the available computational tools and methods to the study of argumentation and develop new tools and methods if needed. More specifically, our project has the following four objectives: (1) To use machine learning tools to learn Bayesian networks from large data sets. (2) To develop adequate argument generation and evaluation algorithms from these Bayesian networks. (3) To set up tools for testing perceived argument quality of generated arguments. (4) To use these tools to test the arguments we generated.