In this work we describe the current stage of development of a software architecture designed to present discrimination tests to pre-school children in the form of gamified tasks. We interpret the problem of administering these tests as a dialogue model using probabilistic rules to generate customised tests on the basis of the child's performance. In the proposed architecture, the dialogue system controls a gaming setup composed of a virtual agent and a robotic companion that needs to be taught how to talk. This learning-by-teaching approach is used to camouflage a phonemes discrimination test that has the added value of being generated at runtime on the basis of the child's performance. We will describe the architectural components involved and we will describe how the dialogue system can make use of linguistic knowledge to generate the discrimination test and administer it by controlling the agents involved in the game.
