Scaling the Burden of Proof to the Cost a Claim Imposes (zenodo.org)

A new research paper introduces the Claim-Offence Theory, which posits that the burden of proof should scale with the potential cost of a claim on individuals. The theory asserts that as the implications of a claim increase, so too does the responsibility of its proponent to provide evidence, while the skeptics bear no obligation to disprove it. The paper evaluates this principle across various contexts, including false accusations, public health declarations, and normative religious doctrines that demand adherence and penalize dissent. This theory is significant for the fields of artificial intelligence and machine learning, particularly in the context of how claims and data-driven assertions are validated or challenged in AI systems. By establishing clear responsibility in evidential requirements, the Claim-Offence Theory could influence the development of algorithms and frameworks that handle disputes over data integrity and model outputs. Co-authored with Claude, an AI system by Anthropic, this paper combines ethical considerations with technical implications, inviting further discussion on the role of AI in establishing and validating claims based on their societal costs.