From Deceptive Outputs to Deceptive Mechanisms: A Causal Framework for Language-Model Deception Research

• URL: http://arxiv.org/abs/2609.04166v1
• Date: Thu, 03 Sep 2026 17:52:19 GMT
• Title: From Deceptive Outputs to Deceptive Mechanisms: A Causal Framework for Language-Model Deception Research
• Abstract summary: We introduce a causal taxonomy separating prior commitment from retrospective report, model preference from realized output, false preference from sensitivity to the utility of misleading a recipient, and deceptive behavior from the provenance of the objective or strategy producing it.<n>We find that deceptive-looking behavior can arise without the corresponding proposed mechanism, while other interventions provide direct evidence that recipient information state can causally affect deceptive preference.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Research and news coverage of language-model deception increasingly attributes human-like mental-state concepts to language models. Such claims can blur the distinction between behavior that looks deceptive and a mechanism that is actually deceptive. We introduce a causal taxonomy separating prior commitment from retrospective report, model preference from realized output, false preference from sensitivity to the utility of misleading a recipient, and deceptive behavior from the provenance of the objective or strategy producing it. We test these distinctions in two open-weight model families. Across controlled guessing-game and stock-trading experiments, we find that deceptive-looking behavior can arise without the corresponding proposed mechanism, while other interventions provide direct evidence that recipient information state can causally affect deceptive preference. These results show that deceptive behavior can provide evidence for a deceptive mechanism. But even evidence for such a mechanism does not establish model agency in the deception.

Related papers

• Evading Chain-of-Thought Monitoring Through Model Poisoning [1.6951239729573633]
  Chain-of-thought (CoT) monitoring is an increasingly important component of AI safety stacks.<n>This work studies the limits of CoT monitoring through the lens of model poisoning.<n> backdoors can be implanted into reasoning models to elicit an attacker-chosen behavior while their CoT traces appear entirely benign.
  arXiv  Detail & Related papers  (2026-08-03T19:32:13Z)
• From Prediction to Justification: Aligning Sentiment Reasoning with Human Rationale via Reinforcement Learning [22.540816588473557]
  ABSA-R1 is a large language model framework designed to mimic the reason-before-predict' cognitive process.<n>We introduce a Cognition-Aligned Reward Model (formerly sentiment-aware reward model) that enforces consistency between the generated reasoning path and the final emotional label.<n>We show that equipping models with this explicit reasoning capability not only enhances interpretability but also yields superior performance in sentiment classification and triplet extraction.
  arXiv  Detail & Related papers  (2026-04-15T01:55:40Z)
• Dissociating Direct Access from Inference in AI Introspection [11.31435294855236]
  Recent work has shown that AI models can introspect.<n>We show that these models detect injected representations via two separable mechanisms.<n>This content-agnostic introspective mechanism is consistent with leading theories in philosophy and psychology.
  arXiv  Detail & Related papers  (2026-03-05T17:39:37Z)
• Large Language Models Report Subjective Experience Under Self-Referential Processing [0.16623291199400023]
  Large language models sometimes produce structured, first-person descriptions that explicitly reference awareness or subjective experience.<n>We investigate one theoretically motivated condition under which such reports arise: self-referential processing.<n>We test whether this regime reliably shifts models toward first-person reports of subjective experience.
  arXiv  Detail & Related papers  (2025-10-27T20:26:30Z)
• Internal Causal Mechanisms Robustly Predict Language Model Out-of-Distribution Behaviors [61.92704516732144]
  We show that the most robust features for correctness prediction are those that play a distinctive causal role in the model's behavior.<n>We propose two methods that leverage causal mechanisms to predict the correctness of model outputs.
  arXiv  Detail & Related papers  (2025-05-17T00:31:39Z)
• Representation Surgery: Theory and Practice of Affine Steering [72.61363182652853]
  Language models often exhibit undesirable behavior, e.g., generating toxic or gender-biased text.<n>One natural (and common) approach to prevent the model from exhibiting undesirable behavior is to steer the model's representations.<n>This paper investigates the formal and empirical properties of steering functions.
  arXiv  Detail & Related papers  (2024-02-15T00:20:30Z)
• Decoding Susceptibility: Modeling Misbelief to Misinformation Through a Computational Approach [61.04606493712002]
  Susceptibility to misinformation describes the degree of belief in unverifiable claims that is not observable. Existing susceptibility studies heavily rely on self-reported beliefs. We propose a computational approach to model users' latent susceptibility levels.
  arXiv  Detail & Related papers  (2023-11-16T07:22:56Z)
• Interpretable Imitation Learning with Dynamic Causal Relations [65.18456572421702]
  We propose to expose captured knowledge in the form of a directed acyclic causal graph. We also design this causal discovery process to be state-dependent, enabling it to model the dynamics in latent causal graphs. The proposed framework is composed of three parts: a dynamic causal discovery module, a causality encoding module, and a prediction module, and is trained in an end-to-end manner.
  arXiv  Detail & Related papers  (2023-09-30T20:59:42Z)
• Robust Transferable Feature Extractors: Learning to Defend Pre-Trained Networks Against White Box Adversaries [69.53730499849023]
  We show that adversarial examples can be successfully transferred to another independently trained model to induce prediction errors. We propose a deep learning-based pre-processing mechanism, which we refer to as a robust transferable feature extractor (RTFE)
  arXiv  Detail & Related papers  (2022-09-14T21:09:34Z)
• Automatic Fake News Detection: Are current models "fact-checking" or "gut-checking"? [0.0]
  Automatic fake news detection models are ostensibly based on logic. It has been shown that these same results, or better, can be achieved without considering the claim at all. This implies that other signals are contained within the examined evidence.
  arXiv  Detail & Related papers  (2022-04-14T21:05:37Z)
• Properties from Mechanisms: An Equivariance Perspective on Identifiable Representation Learning [79.4957965474334]
  Key goal of unsupervised representation learning is "inverting" a data generating process to recover its latent properties. This paper asks, "Can we instead identify latent properties by leveraging knowledge of the mechanisms that govern their evolution?" We provide a complete characterization of the sources of non-identifiability as we vary knowledge about a set of possible mechanisms.
  arXiv  Detail & Related papers  (2021-10-29T14:04:08Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.