You are here: Home TARDIS platform

TARDIS platform


The TARDIS Academic Platform is available here



The TARDIS project aims to develop an open-source platform for online and offline social training for young people at risk of social exclusion, e.g. young people in major urban areas and their suburbs. This platform will be organised around three main components from the user’s point of view :

  • Authoring tools for the definition of training scenarios. They will enable non-computer experts (field practitioners) to specify and describe interactive social training scenarios (an interaction situation between the learner and a virtual agent endowed with a personality) by combining behavioural bricks. Practitioners will be provided with a library of behavioural bricks to construct the specific behaviours of the agent for individual users, based on their knowledge of the users needs. The experts will have the possibility to alter the specific scenarios, to define the agent’s role and personality, and to setup the level of interaction difficulty.
  • Virtual agents with realistic socio-emotional behaviours, designed to support social training in a social and affective serious game. Those agents use text-to-speech, voice and facial expression recognition, in order to adapt to the users’ social attitudes in real-time. The virtual agent makes a credible yet tireless interlocutor. It, therefore provides an ideal way through which young people can train without any social risk. Furthermore, to encourage engagement the training sessions will be follow a serious-game paradigm. The virtual agent will adapt to the user’s actions and social attitudes depending on its personality, and the specific rules of engagement specified by the practitioner for a given scenario. To enable less predictable and more believable behaviours TARDIS’ virtual agents are autonomous, i.e although they follow goals assigned by the pracitioner with respect to the scenario and the training context, they ‘decide’ autonomously which action is best-suited in each situation, without following a predefined script.
  • Behaviour analysis tools. Based on the recorded interactions with the agent, the TARDIS platform will offer high-level analysis of the young individuals’ behaviours and social skills usage by the pracitioners. By the same token, such tools will allow practitioners and therapists with the basis for providing appropriate help to their young trainees. Behaviour analysis derives from the classical social coaching methodology and it is a method currently used by the practitioners at MLVOE. It would be the pracitioner’s responsibility to discuss with the participants their behaviours and how they could improve their social skills in a given context: the TARDIS consortium does not aim to provide a intelligent training environment. However, as will be discussed in the showcase section, the TARDIS consortium does propose to offer an advanced support tool for discussion amongst young trainees and practitioners. Thus, TARDIS combines the expertise of human practitioners for behaviour analysis and coaching with the flexibility and safety of repeatable online game simulations with virtual agents.

For privacy, the interaction sessions will be recorded at the user’s demand only. The recorded data will be available only to the practitioner, and it will be stored only for a limited duration. Appropriate handling of privacy issues is crucial to developing trusting relations between the young people and the practitioners. Following our discussions with the potential users (practitioners, advisers, psychologists and educators), such relations provide the essential basis for successful coaching experience. Throughout the project, the TARDIS consortium will use the following standards: W3C Emotion ML (representation of emotions), SAIBA BML (behaviour), SAIBA FML (communicative function), W3C SMIL (synchronised multi-modal dialogue), HANIM (standard for human body motion), etc.



Grant agreement n° 288578