This paper addresses software plasticity, i.e. the ability of interactive systems to adapt to context of use while preserving user-centered properties. In plasticity, a classical approach consists in concentrating design efforts on a set of pre-defined contexts of use that deserve high quality User Interfaces (UIs), and switching from one to another according to variations of context of use at runtime. However, key contexts of use cannot be finely envisioned at design time, especially when dealing with the specific field of mobility. Thus, we propose a designer's partner tool running on the end-user's mobile device to probe key contexts of use in the wild. The underlying principles are data gathering, bayesian learning, and clustering techniques. Probing key contexts of use can save design efforts.