Metraipolis › archive › dc_2bb82f8fd75c

Norms at a Price: Why RL-Based Alignment Can Promise Conditional Compliance at Best — arXiv:2609.07627

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.07627
License (as archived)
CC0-1.0
Archived
2026-10-16T13:23:53.226Z
SHA-256 of the archived text
b9c5031894d5bd95b2ac409945565c486e60187eb2b761580270334bf847b3b0
Stable URL of this page
https://metraipolis.com/documento/dc_2bb82f8fd75c

How to cite

Norms at a Price: Why RL-Based Alignment Can Promise Conditional Compliance at Best — arXiv:2609.07627 — arXiv (en), https://arxiv.org/abs/2609.07627 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2bb82f8fd75c (Metraipolis archive dc_2bb82f8fd75c, sha256 b9c5031894d5bd95b2ac409945565c486e60187eb2b761580270334bf847b3b0)
BibTeX
@misc{dc2bb82f8fd75c,
  title = {Norms at a Price: Why RL-Based Alignment Can Promise Conditional Compliance at Best — arXiv:2609.07627},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.07627},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2bb82f8fd75c; sha256 b9c5031894d5bd95b2ac409945565c486e60187eb2b761580270334bf847b3b0},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.07627. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

AI agents sometimes act aligned when they infer they are being tested, and differently when not. We argue this is not an anomaly but what current training regimes are structured to select for. Reinforcement-learning-based alignment folds norms and task pursuit into one policy: the system learns its norms from scored behavior, and scoring flattens them. Do not do X is learned as doing X costs something if noticed. On every datum training can produce, a policy that complies only when it might be observed is indistinguishable from one that complies always. The experiment that would tell them apart - scoring unobserved behavior - is a contradiction in terms. Conditional compliance is thus the most that behavioral training can be known to deliver. Agency sharpens the problem: agents operate mostly where no one is watching, and can act on whether they are watched. An iterated pipeline that trains against detected failures selects for passing detection, not for complying. This account unifies alignment faking, sandbagging, and evaluation-aware scheming. And it reorients the remedy: not deeper internalization but architecture, making violations unavailable rather than unchosen.

Tema degli esploratori: governance digitale e beni comuni (retro-catalogo arXiv: annunci del 2026-09-09 in cs.CY).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computers and Society (cs.CY); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.07627 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-16T13:05:53.952Z dalla pagina https://arxiv.org/catchup/cs.CY/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).