Moral Formation Architectures for Artificial Superintelligence: A Research Prospectus
Abstract Contemporary AI governance frameworks focus predominantly on evaluation, monitoring, resilience, auditing, and mitigation after advanced capabilities emerge. While these approaches are important and necessary, they generally begin after powerful capabilities already exist. This prospectus proposes that a developmental layer may be largely absent from current AI governance discussions. Specifically, it explores whether advanced intelligence should undergo structured moral formation during model gestation and development prior to deployment. The paper introduces the concept of moral formation architectures and argues that developmental formation may represent a high-leverage intervention point capable of complementing existing alignment, safety, and governance mechanisms.
Authors
- Charles Dorr (ORCID: https://orcid.org/0009-0003-1342-5526)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-18
- DOI
- https://doi.org/10.5281/zenodo.24354492
- Primary Topic
- Ethics and Social Impacts of AI
- Type
- preprint