Towards Responsible AI: XAI as a Catalyst for Participatory Development Practices

This paper builds on the call for more human-centred XAI (explainable AI), arguing that XAI is a necessary but under-researched component of meaningful participatory AI development: only if a range of different stakeholder communities can understand, interpret, and scrutinise a system, can they meaningfully participate throughout various points of the AI lifecycle. However, while significant research has centred on XAI primarily targetting developers, those with expert knowledge, or specific end-users, XAI solutions tailored to affected communities are currently less explored -- especially how their information needs change along the AI lifecycle and the factors influencing this. By highlighting current challenges and proposing research agenda for moving forward, the paper aims to stimulate the discussion around how XAI can facilitate meaningful involvement from affected communities, as is crucial for the development of more responsible AI systems.