Adaptive Dialogue Systems (ADS) are intelligent systems, able to interact with users via multiple modalities, such as text, speech, gestures, facial expressions and others. Such systems are able to make conversation with their users, usually on a specific, narrow topic. Assistive Living Environments are environments where the users are by definition not competent with technology, due to various factors, such as mental or physical disabilities, injuries, age and others. While technology that helps improve these people’s quality of life exists, many times they cannot access it due to in exible interfaces. ADS, therefore, have the potential to bridge users and technology by acting as a mediator between them. There are several unique challenges posed by this problem, in addition to the challenges faced by a generic ADS. Our contributions to the state of the art focus on Online Dialogue Policy learning which, coupled with other methods we proposed, can lead to an ADS able to exhibit complex behaviour and appear more intelligent. As a consequence, users trust the system more and it becomes more functional as it is able to elicit behavioural information and use it, for example, to make basic diagnoses. We extensively evaluated the proposed algorithms and present our experimental setup and promising results.