A Rights-Based Approach to Trustworthy AI in Social Media

摘要

Social media platforms increasingly use powerful artificial intelligence (AI) that are fed by the vast flows of digital content that may be used to analyze user behavior, mental state, and physical context. New forms of AI-generated content and AI-driven virtual agents present new forms of risks in social media use, the harm of which will be difficult to predict. Delivering trustworthy social media will therefore be increasingly predicated on effectively governing the trustworthiness of its AI components. In this article, we examine different approaches to the governance AI and the Big Data processing that drives it being explored. We identify a potential over-reliance on individual rights at the expense of consideration of collective rights. In response, we propose a collective approach to AI data governance grounded in a legal proposal for universal, non-exclusive data ownership right. We use the Institutional Analysis and Development (IAD) framework to explore the relative costs and benefits on stakeholders in two use cases, one focused on digital content consumers the other focused on digital content knowledge workers. Following an analysis that looks at self-regulation and industry-state co-regulation, we propose governance through shared data ownership. In this way, future social media platforms may be able to maintain trust in their use of AI by committing to no datafication without representation.