Abstract: In the field of automata learning, similar to model-based testing, a target machine is sent interrogative queries with the goal of uncovering its behaviour. The most common algorithm for this in the industry is L*_Mealy, a Mealy-machine-learning variant of Dana Angluin’s 1987 L*. In its current implementation, no means of filtering queries is available. In 2010, however, Fides Aarts and Frits Vaandrager presented a formalism that specifies certain patterns of restriction on the learner: an interface automaton they called the Learning Purpose. Together with it, they devised and implemented framework, including a Mealy machine translator, to learn interface automata, which are similar to I/O automata, where the learner is restricted or rather guided through the learning purpose automaton. In this talk, I am going to explain their learning framework and focus on the learning purpose with the purpose of examining its expressivity in specifying desired patterns in learning, and examining the possibility of extending it.