A new framework called participatory provenance has been developed to audit AI-mediated public consultation summaries for representational accuracy. This method, based on optimal transport theory and causal inference, was applied to Canada's national AI Strategy consultation and revealed that official summaries underperformed a baseline, effectively excluding a significant portion of participants, particularly those expressing dissent or critique. The research also identified factors like brevity and semantic isolation as predictors of representational outcomes, and an accompanying open-source tool aims to enable policymakers to improve summary fidelity. AI
IMPACT Establishes a method to ensure AI summaries of public input are representative, crucial for fair policy-making.
RANK_REASON Academic paper introducing a new framework and methodology. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
