The students are able to:
- apply agreement technologies for analysing and modelling a collaborative multiagent system
- assess the applicability of agreement technologies for human-robot collaboration
- reflect on what is required for human-machine alignment in human-robot collaborative systems
- compare and critique autonomy-centred and interdependence-centred HART methods
- analyse and design a human-robot collaborative system using HART methods
- design a socially intelligent human-robot interaction
- use AI techniques to train robots for intelligent interaction
As robots and other AI systems are getting more and more interwoven into our society and daily lives, it becomes increasingly important that we can effectively collaborate with such systems, and that they are developed in a responsible way. In this course, we address these challenges, providing insight into computational models, design approaches, user experience and connected ELSE aspects of human-robot collaboration and hybrid intelligence.|
The topics of the course include the following:
- Agreement technologies: computational models for regulating and increasing the effectiveness of societies of interacting software agents centred around the notion of ‘agreement’ for example about which course of action to take. Including: normative multiagent systems, argumentation frameworks, and ontology alignment approaches.
- Human-machine alignment: how can we ensure that robots and other AI applications behave in alignment with human and societal needs and values? Including: responsible AI, value alignment, and shared mental models.
- Human-agent/robot teamwork (HART): how can we make automation a team player? Including: autonomy-centred approaches to HART, such as adjustable autonomy and shared control, and coactive design which is centred around human-machine interdependence.
- Social robotics: how to create socially intelligent robots? Including: socially normative robot behaviour, cross-cultural design, user modelling.
- Interactive learning: Design and implement basic reinforcement learning techniques for collaboration policy learning such as reward shaping, policy learning, state compression using deep learning.