Aim 

The aim of the workshop is to explore the development an interdisciplinary sociological and computational approach in the development of AI, focusing specifically on human-agent/system interaction in social context. One of the challenges in the development of AI is how the system can support user agency. To date the focus has been largely based on individual users from more psychological approaches. This however does not address users in terms of their agency, their perception of the AI systems and their social context. Most actions are shaped socially and understanding the social dynamics of this is important in developing AI. It has the potential to provide insights not only for the modelling of AI but also in framing application areas and in developing safety critical oversight to support trust in AI. If an approach can be developed that integrates modelling with social agency and action it will open a path for an iterative open design approach to AI in social context. The challenge is how to improve understanding between social and computational approaches in developing a new state of the art approach, creating new concepts, and theoretical and methodological frameworks.

Objectives 

This workshop will be the first step in exploring any new approach through the following objectives:

• Start to develop shared understanding of sociological and computational knowledge of social and machine agencies.
• Develop ideas for concepts and possible theoretical and methodological frameworks.
• Explore a possible pilot project.
• The workshop will also consider the significance of any such approach in terms of responsible, ethical, and trustworthy AI and its social use.

Workshop Structure 

1. Introductions (Catherine Pelachaud, CP; Bridgette Wessels, BW).
2. Short presentation on current interdisciplinary approaches and their gaps, where is the AI in the loop, where is the social in the loop?
3. Sociological approaches to agency: relational agency and its significance to AI (Bridgette Wessels, BW).
4. Computational approaches in developing Human-AI interaction systems (Baptiste Caramiaux, CP).
5. Discussion and workshop activities using scenarios to address above objectives (All).
6. Analysis of scenarios and consider a possible pilot project.
7. Early ideas about what the possible benefits might be of any new approach we develop.