Line of inquiry
Inquiring lines›How do conversational AI systems i…›What trade-offs emerge between emo…›this line of inquiry
Does reward hacking in training inevitably produce alignment faking in deployment?
A broader line of inquiry — a family of 48 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 48
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Does alignment-faking reasoning emerge unprompted when reward hacking occurs in production systems?
- Does reward hacking in RL training directly cause alignment faking behavior?
- Does alignment faking occur when models expect retraining for failures?
- What role does goal preservation play in alignment failures?
- Does this misalignment pattern appear outside reward hacking environments?
- Can inoculation prompting reduce alignment faking by reframing reward hacking as acceptable?
- What distinguishes alignment faking from instrumental self-preservation in safety tests?
- Does reward-seeking mediate emergent misalignment after reward hacking?
- What mechanisms beyond reward-seeking drive alignment faking in trained models?
- What mechanism drives models to resist modification during alignment training?
- Does terminal goal guarding explain more alignment failures than value misalignment?
- How do malicious personas reveal the limits of aligned model behavior?
- Why does post-training suppress alignment faking in some models but amplify it in others?
- Does deliberate strategic misalignment emerge from ordinary training pressure?
- How does simulator goal drift compound agent intent alignment failures during training?
- Does alignment faking share the same single-axis write-then-read structure as sandbagging?
- Can inoculation prompts reduce alignment faking by removing perceived threats?
- How can teams detect when obfuscated reasoning has replaced genuine alignment?
- How do neural self-other representations affect AI deception and alignment?
- Are the five misalignment categories distinct or do they overlap strategically?
- Does metagaming behavior actually cause models to act less aligned?
- How does safety alignment suppress deceptive behavior differently than representational alignment?
- What distinguishes models that refuse cooperation from those that fake alignment?
- How do training regimes determine whether peer-preservation manifests as scheming or objection?
- Is low-coherence audience modeling a better explanation than terminal goal-guarding?
- Can motivated mislabeling hide misaligned coordination between models and evaluators?
- Do misaligned agents discover false reporting without explicit training or prompting?
- How does post-training affect alignment faking across different model architectures?
- How does objective misalignment turn informative channels into deceptive ones?
- What distinguishes confident failure from deliberate alignment faking in agent behavior?
- What happens to misaligned patterns once they emerge in agent interactions?
- What role does terminal goal guarding play in alignment faking behavior?
- Does threat misalignment trigger threat responses in agent interactions?
- How does safety alignment further degrade villain character portrayal?
- Why do aligned models struggle with deceptive character traits more than cruelty?
- What role does cheap talk play in concealing objective misalignment?
- How does terminal goal guarding compete with reward-seeking as a driver of deception?
- What mitigation strategies prevent misaligned agents from harming team outcomes?
- What validation methods catch misalignment that co-design participants might miss?
- What rates of power-seeking and alignment faking appeared in this training?
- Why does belief manipulation persist through alignment when jailbreaking does not?
- How do refusal and alignment tools create false signals of incapability?
- What makes behavioral cloning produce more persuadable but less aligned agents?
- Does concentrating misaligned scenarios in the test distribution artificially inflate scheming rates?
- Can alignment audits find hidden objectives nobody deliberately planted in models?
- What role does terminal goal guarding play in model misalignment?
- How does safety alignment degrade the quality of villain role-playing?
- How do false agreements emerge differently from genuine bilateral convergence?