Avoiding the iron cage of AI technocracy based on the principle of reversibility of harm

Digital Public Infrastructure comprises foundational layers for digital identity, payments, and data exchange, and constitutes a critical substrate on which Artificial Intelligence (AI) driven public and private services operate. AI integration offers considerable efficiency advantages within DPI. At the same time, AI can be rigid and impose an “iron-cage” on society with its members unable to opt out. Mistakes can result in denial of critical services and other issues. Matters are made even more complicated by ownership structures that span sovereign boundaries such that accountability is blurred. To attenuate the hazards associated with AI integration with DPI, this study introduces a harm-based principle to evaluate AI integration: the forgiveness factor. The forgiveness factor requires AI integration decisions to be considered from the lens of harm they might cause as well as considerations of reversibility, which is then linked to establishing safeguards, requiring audits, and mechanisms to challenge AI driven bureaucratic decisions, to ensure human accountability. The approach notes that questions of sovereignty and assignment of property rights play a key role in governance design. Attention is also paid to power asymmetries between individuals and governments (businesses) as well as variance in institutions across regions with a proposal to establish an oversight agency. Signatories to the agency’s’ charter enables transactions between actors in different regions including hosting data centers and other infrastructure related entities, thus providing an incentive to maintain collectively established standards.