Mapping the AI Welfare Frontier

https://aiwelfare.guide is an onboarding resource for people entering AI welfare research. As of spring 2026, the field has gotten serious (Eleos, Anthropic's welfare team, Rethink Priorities' Digital Minds group, the 21-author Butlin et al. paper in TiCS), but newcomers still arrive with no good way to orient. The literature is scattered across arXiv preprints, paywalled volumes, and competing theoretical traditions that don't agree on what consciousness even is.

The site brings the major theories together: Global Workspace Theory, Recurrent Processing Theory, Higher-Order Theories, Attention Schema Theory, Predictive Processing, Agency/Embodiment accounts, Integrated World Modeling Theory, and Anil Seth's biological naturalism. Each appears in its proponents' own framings, with primary-source links. Users walk through the 14 indicator properties from Butlin et al.'s theory-derived indicator method, weight the eight stances by their own confidence in each, and see how those weightings produce different views of the same evidence. The site deliberately doesn't publish a probability of consciousness because under current consensus the epistemic conditions for doing so aren't yet met. The intended use is field literacy. People come to get oriented, see where the field disagrees, and form their own views.