Computational virtue ethics framework

Cosmos Institute granted Rocco Hu on September 10, 2024 as part of the inaugural Cosmos Ventures cohort, a set of small grants ($1K to $10K range) designed to incubate prototypes at the intersection of philosophy and AI. Cosmos Institute's stated remit, articulated when it launched alongside an AI lab with Anthropic co-founder Jack Clark as a founding fellow, is backing builders working on reason, decentralization, and human autonomy in AI.

Hu is an Oxford-affiliated philosopher and computer scientist whose declared interests in metaphysics, epistemology, and value theory are listed on his PhilPeople profile. The project develops a computational theory of virtue ethics intended to ground AI value learning in an Aristotelian rather than utilitarian frame. The Cosmos Institute's own framing of the grant describes the work as "developing a computational theory of virtue ethics to embed ethical reasoning in both human and AI systems."

The technical question the framing implies is whether virtue-ethics primitives (habituation, phronesis or practical wisdom, character traits as stable dispositions rather than per-decision utility calculations) can be operationalized inside model training and evaluation, rather than treated only as commentary on alignment from outside the loop. This contrasts with the dominant alignment frames of RLHF and constitutional approaches, which encode rules or preferences at the policy level rather than modeling character as a stable trait acquired through structured habituation. The project sits at the evaluation and governance meta-layers: virtue ethics as a frame produces different benchmark questions (what dispositions does the model exhibit under perturbation) than utilitarian frames (what outcomes does it maximize).