AI RESEARCH

Adaptive Pluralistic Alignment: A pipeline for dynamic artificial democracy

arXiv CS.LG

ArXi:2605.01642v1 Announce Type: new Prevailing alignment methods target a fixed set of preferences and therefore risk forcing value lock-in as societal norms evolve over time. We