arxivcs.CYcs.AIcs.CLcs.LG2026-07-18
A Method for Learning Value Systems in Generative AI
Andrés Holgado-Sánchez, Holger Billhardt, Sascha Ossowski
Value-aware AI systems require explicit computational representations of human values (groundings) and their aggregation into value systems in order to align their decisions with ours. As such representations are difficult to elicit, value learning seeks to infer them by observin…