Policy1d ago
Policy
Applying Shapley Values To AI Responsibility
A new proposal on LessWrong suggests using Shapley shares to quantify responsibility for second-order effects. This framework treats causal contributions as marginal additions to a collective outcome. While theoretical, it offers a mathematical approach to attributing blame in complex systems. Practitioners can use this to model accountability in multi-agent AI deployments.