Data stewards in service of Artificial Intelligence: Reimagining AI futures towards a participatory paradigm for technological innovation

 The successful development and deployment of AI systems depends on access to data which is used to train models using several different techniques - from machine learning to automation. However, issues related to underlying datasets which are used to train algorithms and bottlenecks within organisations which undertake AI development result in AI-driven products or services that fail to scale due to concerns regarding bias, quality and unfair use of data. Who has access to the data used to build AI systems, what are the conditions under which the data is shared and who benefits from data use are some of the significant questions that remain unaddressed under prevailing logics of AI research and development. These systemic issues combined with prevailing power asymmetries in AI research and development result in arbitrary exclusion of individuals and communities - who are the primary producers of data - from participating in algorithmic governance and decision making. Regulation on how AI is researched and developed requires a paradigm change to push for responsible AI. Institutional frameworks for regulation of AI should adopt perspectives from procedural justice praxis to ensure that fundamental human rights are upheld and create space for public dialogue around AI deployment for specific contexts and purposes. It is this paper’s contention that embedding data stewardship - an approach to data governance which unlocks data for responsible use without compromising the agency of individuals and communities that produce the data - can go a long way in advancing AI innovation through safe, trustworthy and fair mechanisms.