Leibowicz study: Synthetic media disclosures suffer from normativity vs neutrality tension

A new study by Claire Leibowicz, published on arXiv, dives into the upstream decision-making of AI policymakers and practitioners tasked with designing disclosure labels for synthetic media. Based on 23 expert interviews and 13 case studies from organizations in the Partnership on AI's Synthetic Media Framework, the research identifies two core disclosure goals: process transparency and harm reduction. However, these goals clash with two central tensions—normativity versus neutrality and proactivity versus precision. Policymakers struggle to decide whether labels should judge content (normative) or merely describe it (neutral), and whether to warn proactively or wait for precise detection.

The study reveals that analogical reasoning—comparing synthetic media disclosures to nutrition labels or California's Proposition 65 warnings—is a common strategy to navigate these tensions, but it often masks underlying conflicts rather than resolving them. Leibowitz argues that the analogies themselves shape design choices in ways that may mislead audiences. The paper calls for more scholarship focused on the decision-makers behind AI transparency tools, noting that the design of disclosures will ultimately determine how billions of users evaluate media credibility in the age of generative AI.