Socialt skratt för virtuella agenter

Laughter is not only a reaction to a joke, but plays a crucial role in managing miscommunication in dialogue. We hypothesise that theoretical models of laughter in human dialogue extends to virtual entities.

We aim to confirm this using experimental methods which involve virtual agents that have a body and are able to communicate both verbally and non-verbally.

SocLaVA will develop and evaluate virtual agents which produce laughter behaviours that are appropriate in socially incongruous contexts. Modelling conversational competence computationally is a key challenge in cognitive science research.

Theoretical and computational models of laughter are derived from human conversations and haven’t yet been tested in interactions with virtual agents.

SocLaVA aims to scrutinise and improve current theoretical models, and provide more evidence for future theoretical contributions.SocLaVA will produce experimental data which will reflect perception of social laughter in interactions with virtual agents.

We will use several setups with different levels of immersion – agents on-screen, in virtual reality and embodied in a robotic head.

Virtual agents will be evaluated in terms of user perception of agent along the social dimensions of warmth and competence. Quality of interactions will be assesed.

The project will contribute to formal and computational approaches laughter in interaction and the development of truly conversational embodied agent.