Social robots operate in a world that includes humans, leading to highly complex tasks. For example, if a robot has to navigate towards a person to interact with, it needs to plan an approach that is acceptable for both the person it is navigating towards, and the people it passes on its way.
Recently, advances have been made in social robotics. For example, by combining insights from psychology and sociology, interaction data, and computer vision and machine learning algorithms, it is possible to learn socially acceptable control policies from data. This opens the door to general approaches to creating socially acceptable policies, without having to handcraft them. However, because data is typically very sparse, and depends on human input, this remains a challenging problem.In this workshop we will focus on the future of social robotics, i.e., we look at what is already achieved and discuss important open questions, such as: how can we learn from small amounts of human interaction data; how can we efficiently measure performance empirically; how can we get sufficient feedback from humans; how should we interpret that feedback; how should we model human intentions; how can we give social robots good morals; and what ethical dilemmas might a social robot face?
The workshop will take place in Lecture Theatre A (LTA) on 2 June 2016 from 13:30-17:30, at the Department of Computer Science, University of Oxford, Wolfson Building, Parks Road, Oxford OX1 3QD (UK).
The workshop is free, and open to anyone who wishes to attend.
Provisional planning:
This workshop is sponsored by the TERESA project.
