Fair outputs, Biased Internals: Causal Potency and Asymmetry of Latent Bias in LLMs for High-Stakes Decisions

• URL: http://arxiv.org/abs/2605.15217v1
• Date: Tue, 12 May 2026 12:14:58 GMT
• Title: Fair outputs, Biased Internals: Causal Potency and Asymmetry of Latent Bias in LLMs for High-Stakes Decisions
• Abstract summary: We investigate the use of open-weight models for mortgage underwriting using matched applications that differ only in racially-associated names.<n>We show that models show no output-level bias, yet retain and amplify demographic representations across model layers.<n>Through activation steering and novel cross-layer interventions, we demonstrate that this suppressed information is decision-relevant.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Instruction-tuned language models exhibit behavioural fairness in high-stakes decisions while retaining biased associations in their internal representations. However, whether these suppressed representations can affect model outputs - and whether such causal potency is symmetric across demographic groups - remains unknown. We investigate the use of open-weight models for mortgage underwriting using matched applications that differ only in racially-associated names and reveal a critical disconnect: models show no output-level bias, yet retain and amplify demographic representations across model layers. Through activation steering and novel cross-layer interventions, we demonstrate that this suppressed information is decision-relevant: when reinjected at critical layers, it produces near-complete decision reversals. Critically, this latent bias is asymmetric - steering interventions affect decisions in one demographic direction, while producing minimal effects in reverse - and susceptible to adversarial prompt engineering and parameter-efficient fine-tuning. These findings demonstrate that behavioural audits focused on outputs are insufficient: fair outputs can mask exploitable internal biases. They also motivate dual-layer testing frameworks combining output evaluation with representational analysis for AI governance in high-stakes decisions.

Related papers

• The Illusion of Intervention: Your LLM-Simulated Experiment is an Observational Study [56.649987197786096]
  Large language models (LLMs) show potential as simulators of human behavior.<n>LLMs can induce unintended shifts in latent user attributes, causing user drift.<n>We formalize the confounding or selection bias that can arise due to user drift.
  arXiv  Detail & Related papers  (2026-05-20T06:09:41Z)
• Same Answer, Different Representations: Hidden instability in VLMs [65.36933543377346]
  We introduce a representation-aware and frequency-aware evaluation framework that measures internal embedding drift, spectral sensitivity, and structural smoothness.<n>We apply this framework to modern Vision Language Models (VLMs) across the SEEDBench, MMMU, and POPE datasets.
  arXiv  Detail & Related papers  (2026-02-06T12:24:26Z)
• DSO: Direct Steering Optimization for Bias Mitigation [12.033608044339717]
  Generative models are often deployed to make decisions on behalf of users, such as vision-language models (VLMs) identifying which person in a room is a doctor to help visually impaired individuals.<n>Yet, VLM decisions are influenced by the perceived demographic attributes of people in the input, which can lead to biased outcomes like failing to identify women as doctors.<n>We propose Direct Steering Optimization (DSO) which uses reinforcement learning to find linear transformations for steering activations, tailored to mitigate bias while maintaining control over model performance.
  arXiv  Detail & Related papers  (2025-12-17T19:43:46Z)
• Decomposing Direct and Indirect Biases in Linear Models under Demographic Parity Constraint [4.129225533930966]
  We propose a post-processing framework to decompose the resulting bias into direct (sensitive-attribute) and indirect (correlated-features) components.<n>Our method analytically characterizes how demographic parity reshapes each model coefficient, including those of both sensitive and non-sensitive features.<n>Our framework requires no retraining and provides actionable insights for model auditing and mitigation.
  arXiv  Detail & Related papers  (2025-11-14T13:27:54Z)
• Silenced Biases: The Dark Side LLMs Learned to Refuse [5.2630646053506345]
  We introduce the concept of silenced biases, which are unfair preferences encoded within models' latent space.<n>We propose the Silenced Bias Benchmark (SBB), which aims to uncover these biases by employing activation steering.
  arXiv  Detail & Related papers  (2025-11-05T11:24:50Z)
• Self-Adaptive Cognitive Debiasing for Large Language Models in Decision-Making [71.71796367760112]
  Large language models (LLMs) have shown potential in supporting decision-making applications.<n>We propose a cognitive debiasing approach, self-adaptive cognitive debiasing (SACD)<n>We evaluate SACD on finance, healthcare, and legal decision-making tasks using both open-weight and closed-weight LLMs.
  arXiv  Detail & Related papers  (2025-04-05T11:23:05Z)
• Achieving Group Fairness through Independence in Predictive Process Monitoring [0.0]
  Predictive process monitoring focuses on forecasting future states of ongoing process executions, such as predicting the outcome of a particular case.<n>In recent years, the application of machine learning models in this domain has garnered significant scientific attention.<n>This work addresses group fairness in predictive process monitoring by investigating independence, i.e. ensuring predictions are unaffected by sensitive group membership.
  arXiv  Detail & Related papers  (2024-12-06T10:10:47Z)
• Identifying and Mitigating Social Bias Knowledge in Language Models [52.52955281662332]
  We propose a novel debiasing approach, Fairness Stamp (FAST), which enables fine-grained calibration of individual social biases.<n>FAST surpasses state-of-the-art baselines with superior debiasing performance.<n>This highlights the potential of fine-grained debiasing strategies to achieve fairness in large language models.
  arXiv  Detail & Related papers  (2024-08-07T17:14:58Z)
• Hierarchical Bias-Driven Stratification for Interpretable Causal Effect Estimation [1.6874375111244329]
  BICauseTree is an interpretable balancing method that identifies clusters where natural experiments occur locally. We evaluate the method's performance using synthetic and realistic datasets, explore its bias-interpretability tradeoff, and show that it is comparable with existing approaches.
  arXiv  Detail & Related papers  (2024-01-31T10:58:13Z)
• 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)
• Delving into Identify-Emphasize Paradigm for Combating Unknown Bias [52.76758938921129]
  We propose an effective bias-conflicting scoring method (ECS) to boost the identification accuracy. We also propose gradient alignment (GA) to balance the contributions of the mined bias-aligned and bias-conflicting samples. Experiments are conducted on multiple datasets in various settings, demonstrating that the proposed solution can mitigate the impact of unknown biases.
  arXiv  Detail & Related papers  (2023-02-22T14:50:24Z)
• Deconfounding Scores: Feature Representations for Causal Effect Estimation with Weak Overlap [140.98628848491146]
  We introduce deconfounding scores, which induce better overlap without biasing the target of estimation. We show that deconfounding scores satisfy a zero-covariance condition that is identifiable in observed data. In particular, we show that this technique could be an attractive alternative to standard regularizations.
  arXiv  Detail & Related papers  (2021-04-12T18:50:11Z)

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.