HARPI: Anonymous research for human survival

Field notes

HARPI divides its stated program among advanced-AI alignment and evaluation, mathematical work on uncertainty and decision-making, and policy analysis of institutions and incentives. Its homepage also sets out an editorial method: make assumptions explicit, separate evidence from judgment, and leave conclusions open to revision. The institute invites original research, reviews, and policy collaborations, while its authorship policy permits anonymous contributions. A featured letter to Congress proposes an “Absolute Alignment Act.”