Proves that no behavior-dependent feedback training strategy can guarantee an honest agent for latent knowledge even with perfect training feedback.
Propositional interpretability in artificial intelligence
6 Pith papers cite this work, alongside 2 external citations. Polarity classification is still indexing.
6
Pith papers citing it
2
external citations · external index
verdicts
UNVERDICTED 6representative citing papers
A framework is proposed for solving for an AI system's beliefs and desires from its computational facts, with criteria for success tied to interpretability tests and emphasis on holistic attribution.
Off-policy training data for LLM behavior probes causes significant generalization failures especially for intent-based behaviors like deception, and performance on coerced incentivised data correlates with real on-policy success.
Persona axes derived from contrastive prompts and PCA yield linear probes that generalize better than raw-activation probes across 10 datasets for deception and sycophancy.
Supervised learning across AI systems vindicates a uniform error-driven associationism for cognition, though operating inside advanced computational structures beyond classical associationist models.
The paper claims that mechanistic interpretability needs philosophy as a partner to clarify concepts, refine methods, and navigate epistemic and ethical complexities in AI systems.
citing papers explorer
-
The Impossibility of Eliciting Latent Knowledge
Proves that no behavior-dependent feedback training strategy can guarantee an honest agent for latent knowledge even with perfect training feedback.
-
Radical AI Interpretability
A framework is proposed for solving for an AI system's beliefs and desires from its computational facts, with criteria for success tied to interpretability tests and emphasis on holistic attribution.
-
The Impact of Off-Policy Training Data on Probe Generalisation
Off-policy training data for LLM behavior probes causes significant generalization failures especially for intent-based behaviors like deception, and performance on coerced incentivised data correlates with real on-policy success.
-
Do Linear Probes Generalize Better in Persona Coordinates?
Persona axes derived from contrastive prompts and PCA yield linear probes that generalize better than raw-activation probes across 10 datasets for deception and sycophancy.
-
The New Associationism: Lessons from Deep Learning
Supervised learning across AI systems vindicates a uniform error-driven associationism for cognition, though operating inside advanced computational structures beyond classical associationist models.
-
Mechanistic Interpretability Needs Philosophy
The paper claims that mechanistic interpretability needs philosophy as a partner to clarify concepts, refine methods, and navigate epistemic and ethical complexities in AI systems.