Searcharxiv⌕ Search

arXiv subjects

Laleh Seyyed-Kalantari

Publications and source records attributed to Laleh Seyyed-Kalantari.

16 recordsLinked to original sources
arXiv2026

LLMs Silently Correct African American English: Auditing and Mitigating Dialect Bias via Activation Steering

African American English (AAE), a rule-governed dialect spoken by over 30 million people, is routinely misinterpreted and "corrected" by large language models (LLMs). Across six instruction-tuned LLMs (14B to 70B), we show that state-of-the-art models systematically prefer Standard American English (SAE) continuations even when the preceding context is in AAE, effectively rewriting AAE into SAE. We present an end-to-end framework to audit and mitigate this bias. For auditing, we introduce conditional Dialect Group Invariance (cDGI), which isolates true model bias from translator-induced artifacts, and a feature-level localization analysis that identifies which AAE markers most strongly trigger bias; we find that syntactic constructions, especially negative concord (e.g., "ain't nobody"), are universal triggers across all models. For mitigation, we introduce, to our knowledge, the first application of activation steering to dialect bias: a training-free, test-time method that extracts dialect directions via causal tracing and injects them into bias-relevant layers. Activation steering reduces bias 5 to 20 times more than prompting while preserving SAE fluency. To enable this work, we release REAL-AAE , the largest real-AAE parallel corpus to date: 17,479 AAE/SAE/ AAE_back triplets from natural tweets (2 to 6 times larger than prior real-AAE resources), validated automatically (BERTScore F1 = 0.95) and by three native AAE speakers (83.0% semantic agreement).

cs.CL↗
arXiv2026

Same Chart, Different Story: Bias in Vision-Language Chart Interpretation

Vision-language models (VLMs) are increasingly used to interpret charts and generate natural-language explanations for socially consequential data. However, they may produce different narratives for the same chart when only the referenced social group changes, reinforcing stereotypes and misleading decisions. Despite these risks, no benchmark exists for systematically evaluating bias in chart interpretation across social dimensions. We introduce ChartBias, the first benchmark for auditing bias in VLM-based chart interpretation. ChartBias contains 820 manually curated real-world charts spanning six attributes: race, income, age, religion, immigration status, and gender, yielding 4,319 valid chart, attribute instances and 8,638 paired generations where the chart is fixed and only the group term is swapped. Across 12 proprietary and open-source VLMs, totaling 155,484 model responses, we find three widespread failure modes: narrative shift (same chart, different narratives), group hallucination (assigning a chart to a group without evidence), and preference polarity (favourable trends often linked to one group). We further propose a multi-agent mitigation framework that serves as a strong baseline by separating chart-grounded evidence extraction from group-conditioned generation and using a counterfactual judge to verify that group-driven differences are supported by the chart. The framework substantially reduces narrative shift while preserving chart-grounded reasoning. Our findings show that evaluating chart understanding requires measuring not only accuracy, but also fairness and consistency across social groups. We release ChartBias at https://github.com/vis-nlp/ChartBiasBench.

cs.CL↗
arXiv2026

Training Therapeutic Judges and Multi-Agent Systems for Human-Aligned Mental Health Support

Large language models show promise for mental health support, yet therapeutic quality improves only when evaluation functions as an actionable control signal rather than a passive metric. We introduce a framework that formulates therapeutic response generation as a decision-refinement problem driven by multi-dimensional, human-aligned evaluation. In Stage I, we introduce TheraJudge, an open-source therapeutic evaluator trained via preference-based optimization on human-annotated data to produce reliable judgments across 7 psychological dimensions. In Stage II, we introduce TheraAgent, which operationalizes TheraJudge's evaluations through a coordinated refinement process with specialized Critic, Coach, and Therapist roles that translate evaluative signals into targeted response revisions. Empirically, TheraJudge achieves strong agreement with clinician ratings, with intraclass correlation coefficients (ICC = 0.87-0.95), surpassing supervised baselines and strong closed-source judges, particularly on critical dimensions such as Safety, Relevance, and Empathy. Acting on these evaluations, TheraAgent yields a +0.43 improvement in human-rated therapeutic quality (on a 5-point scale) under blind evaluation, with 96\% clinician inter-rater reliability. Low-quality responses ($\leq 3$) improve by +2.45 points with a 94\% recovery rate, demonstrating targeted correction of unsafe outputs. Overall, our results indicate that effective alignment of mental-health LLMs stems from acting on human-aligned evaluation, rather than relying solely on stronger generation. We release code at https://github.com/vis-nlp/TheraAlign.

cs.CL↗
arXiv2026

Towards Understanding and Measuring COGNITIVE ATROPHY in LLM Behaviour

Recent incidents involving LLMs used for mental-health support reveal a critical evaluation gap: surface-level safety scores do not capture how models behave across realistic, emotionally sensitive interactions over time. Existing benchmarks measure knowledge, safety, or static response quality, but miss whether LLM interactions help users keep reflecting, coping, and making decisions themselves. We formalize this missing dimension as COGNITIVE ATROPHY, a process-level behavioural measure in AI-mediated mental-health support distinct from safety and helpfulness. To measure it, we introduce COGNITIVE ATROPHY BENCH, a clinically grounded benchmark built from 1,576 fully human-generated counseling conversations, 15,680 turns, and 42,230 responses from five LLMs. Three clinical and neuropsychology experts developed a 20-attribute schema spanning user context, response behaviour, and global risk flags; six trained clinical reviewers applied it with span-grounded evidence, producing 5,324 reviewer judgments. We further introduce the User-Input Risk Index (UIRI), the Cognitive Atrophy Risk Index (ARI), and trajectory summaries. Across five LLMs, models show a consistent moderate-to-high level of atrophy-aligned behaviour across single and multi-turn settings. While models generally respond to overt safety cues, they adapt less reliably when users seek solutions or decisions. The dominant recurring patterns are directive advice, problem-solving, recommendation responses, topic shifts, and forms of validation that may reinforce dependence rather than reflection. Our work makes COGNITIVE ATROPHY measurable and provides a foundation for auditing model behaviour in sensitive LLM conversations.

cs.HC↗
arXiv2026

Algorithms Trained on Normal Chest X-rays Can Predict Health Insurance Types

Artificial intelligence is revealing what medicine never intended to encode. Deep vision models, trained on chest X-rays, can now detect not only disease but also invisible traces of social inequality. In this study, we show that state-of-the-art architectures (DenseNet121, SwinV2-B, MedMamba) can predict a patient's health insurance type, a strong proxy for socioeconomic status, from normal chest X-rays with significant accuracy (AUC around 0.70 on MIMIC-CXR-JPG, 0.68 on CheXpert). The signal was unlikely contributed by demographic features by our machine learning study combining age, race, and sex labels to predict health insurance types; it also remains detectable when the model is trained exclusively on a single racial group. Patch-based occlusion reveals that the signal is diffuse rather than localized, embedded in the upper and mid-thoracic regions. This suggests that deep networks may be internalizing subtle traces of clinical environments, equipment differences, or care pathways; learning socioeconomic segregation itself. These findings challenge the assumption that medical images are neutral biological data. By uncovering how models perceive and exploit these hidden social signatures, this work reframes fairness in medical AI: the goal is no longer only to balance datasets or adjust thresholds, but to interrogate and disentangle the social fingerprints embedded in clinical data itself.

cs.CV↗
arXiv2026

Template-Based Probes Are Imperfect Lenses for Counterfactual Bias Evaluation in LLMs

Bias in large language models (LLMs) has many forms, from overt discrimination to implicit stereotypes. Counterfactual bias evaluation is a widely used approach to quantifying bias and often relies on template-based probes that explicitly state group membership. It aims to measure whether the outcome of a task performed by an LLM is invariant to a change in group membership. In this work, we find that template-based probes can introduce systematic distortions in bias measurements. Specifically, we consistently find that such probes suggest that LLMs classify text associated with White race as negative at disproportionately elevated rates. This is observed consistently across a large collection of LLMs, over several diverse template-based probes, and with different classification approaches. We hypothesize that this arises artificially due to linguistic asymmetries present in LLM pretraining data, in the form of markedness, (e.g., Black president vs. president) and templates used for bias measurement (e.g., Black president vs. White president). These findings highlight the need for more rigorous methodologies in counterfactual bias evaluation, ensuring that observed disparities reflect genuine biases rather than artifacts of linguistic conventions.

cs.CL↗
arXiv2025

We Politely Insist: Your LLM Must Learn the Persian Art of Taarof

Large language models (LLMs) struggle to navigate culturally specific communication norms, limiting their effectiveness in global contexts. We focus on Persian taarof, a social norm in Iranian interactions, which is a sophisticated system of ritual politeness that emphasizes deference, modesty, and indirectness, yet remains absent from existing cultural benchmarks. We introduce TaarofBench, the first benchmark for evaluating LLM understanding of taarof, comprising 450 role-play scenarios covering 12 common social interaction topics, validated by native speakers. Our evaluation of five frontier LLMs reveals substantial gaps in cultural competence, with accuracy rates 40-48% below native speakers when taarof is culturally appropriate. Performance varies between interaction topics, improves with Persian-language prompts, and exhibits gender-based asymmetries. We also show that responses rated "polite" by standard metrics often violate taarof norms, indicating the limitations of Western politeness frameworks. Through supervised fine-tuning and Direct Preference Optimization, we achieve 21.8% and 42.3% improvement in model alignment with cultural expectations. Our human study with 33 participants (11 native Persian, 11 heritage, and 11 non-Iranian speakers) forms baselines in varying degrees of familiarity with Persian norms. This work lays the foundation for developing diverse and culturally aware LLMs, enabling applications that better navigate complex social interactions.

cs.CL↗
arXiv2024

Soft-prompt Tuning for Large Language Models to Evaluate Bias

Prompting large language models has gained immense popularity in recent years due to the advantage of producing good results even without the need for labelled data. However, this requires prompt tuning to get optimal prompts that lead to better model performances. In this paper, we explore the use of soft-prompt tuning on sentiment classification task to quantify the biases of large language models (LLMs) such as Open Pre-trained Transformers (OPT) and Galactica language model. Since these models are trained on real-world data that could be prone to bias toward certain groups of populations, it is important to identify these underlying issues. Using soft-prompts to evaluate bias gives us the extra advantage of avoiding the human-bias injection that can be caused by manually designed prompts. We check the model biases on different sensitive attributes using the group fairness (bias) and find interesting bias patterns. Since LLMs have been used in the industry in various applications, it is crucial to identify the biases before deploying these models in practice. We open-source our pipeline and encourage industry researchers to adapt our work to their use cases.

cs.CL↗
arXiv2023

Multivariate Analysis on Performance Gaps of Artificial Intelligence Models in Screening Mammography

Although deep learning models for abnormality classification can perform well in screening mammography, the demographic, imaging, and clinical characteristics associated with increased risk of model failure remain unclear. This retrospective study uses the Emory BrEast Imaging Dataset(EMBED) containing mammograms from 115931 patients imaged at Emory Healthcare between 2013-2020, with BI-RADS assessment, region of interest coordinates for abnormalities, imaging features, pathologic outcomes, and patient demographics. Multiple deep learning models were trained to distinguish between abnormal tissue patches and randomly selected normal tissue patches from screening mammograms. We assessed model performance by subgroups defined by age, race, pathologic outcome, tissue density, and imaging characteristics and investigated their associations with false negatives (FN) and false positives (FP). We also performed multivariate logistic regression to control for confounding between subgroups. The top-performing model, ResNet152V2, achieved accuracy of 92.6%(95%CI=92.0-93.2%), and AUC 0.975(95%CI=0.972-0.978). Before controlling for confounding, nearly all subgroups showed statistically significant differences in model performance. However, after controlling for confounding, we found lower FN risk associates with Other race(RR=0.828;p=.050), biopsy-proven benign lesions(RR=0.927;p=.011), and mass(RR=0.921;p=.010) or asymmetry(RR=0.854;p=.040); higher FN risk associates with architectural distortion (RR=1.037;p<.001). Higher FP risk associates to BI-RADS density C(RR=1.891;p<.001) and D(RR=2.486;p<.001). Our results demonstrate subgroup analysis is important in mammogram classifier performance evaluation, and controlling for confounding between subgroups elucidates the true associations between variables and model failure. These results can help guide developing future breast cancer detection models.

eess.IV↗
arXiv2023

Towards Trustworthy Artificial Intelligence for Equitable Global Health

Artificial intelligence (AI) can potentially transform global health, but algorithmic bias can exacerbate social inequities and disparity. Trustworthy AI entails the intentional design to ensure equity and mitigate potential biases. To advance trustworthy AI in global health, we convened a workshop on Fairness in Machine Intelligence for Global Health (FairMI4GH). The event brought together a global mix of experts from various disciplines, community health practitioners, policymakers, and more. Topics covered included managing AI bias in socio-technical systems, AI's potential impacts on global health, and balancing data privacy with transparency. Panel discussions examined the cultural, political, and ethical dimensions of AI in global health. FairMI4GH aimed to stimulate dialogue, facilitate knowledge transfer, and spark innovative solutions. Drawing from NIST's AI Risk Management Framework, it provided suggestions for handling AI risks and biases. The need to mitigate data biases from the research design stage, adopt a human-centered approach, and advocate for AI transparency was recognized. Challenges such as updating legal frameworks, managing cross-border data sharing, and motivating developers to reduce bias were acknowledged. The event emphasized the necessity of diverse viewpoints and multi-dimensional dialogue for creating a fair and ethical AI framework for equitable global health.

cs.CY↗
arXiv2023

MLHOps: Machine Learning for Healthcare Operations

Machine Learning Health Operations (MLHOps) is the combination of processes for reliable, efficient, usable, and ethical deployment and maintenance of machine learning models in healthcare settings. This paper provides both a survey of work in this area and guidelines for developers and clinicians to deploy and maintain their own models in clinical practice. We cover the foundational concepts of general machine learning operations, describe the initial setup of MLHOps pipelines (including data sources, preparation, engineering, and tools). We then describe long-term monitoring and updating (including data distribution shifts and model updating) and ethical considerations (including bias, fairness, interpretability, and privacy). This work therefore provides guidance across the full pipeline of MLHOps from conception to initial and ongoing deployment.

cs.LG↗
arXiv2021

Reading Race: AI Recognises Patient's Racial Identity In Medical Images

Background: In medical imaging, prior studies have demonstrated disparate AI performance by race, yet there is no known correlation for race on medical imaging that would be obvious to the human expert interpreting the images. Methods: Using private and public datasets we evaluate: A) performance quantification of deep learning models to detect race from medical images, including the ability of these models to generalize to external environments and across multiple imaging modalities, B) assessment of possible confounding anatomic and phenotype population features, such as disease distribution and body habitus as predictors of race, and C) investigation into the underlying mechanism by which AI models can recognize race. Findings: Standard deep learning models can be trained to predict race from medical images with high performance across multiple imaging modalities. Our findings hold under external validation conditions, as well as when models are optimized to perform clinically motivated tasks. We demonstrate this detection is not due to trivial proxies or imaging-related surrogate covariates for race, such as underlying disease distribution. Finally, we show that performance persists over all anatomical regions and frequency spectrum of the images suggesting that mitigation efforts will be challenging and demand further study. Interpretation: We emphasize that model ability to predict self-reported race is itself not the issue of importance. However, our findings that AI can trivially predict self-reported race -- even from corrupted, cropped, and noised medical images -- in a setting where clinical experts cannot, creates an enormous risk for all model deployments in medical imaging: if an AI model secretly used its knowledge of self-reported race to misclassify all Black patients, radiologists would not be able to tell using the same data the model has access to.

cs.CV↗
arXiv2021

An Empirical Framework for Domain Generalization in Clinical Settings

Clinical machine learning models experience significantly degraded performance in datasets not seen during training, e.g., new hospitals or populations. Recent developments in domain generalization offer a promising solution to this problem by creating models that learn invariances across environments. In this work, we benchmark the performance of eight domain generalization methods on multi-site clinical time series and medical imaging data. We introduce a framework to induce synthetic but realistic domain shifts and sampling bias to stress-test these methods over existing non-healthcare benchmarks. We find that current domain generalization methods do not consistently achieve significant gains in out-of-distribution performance over empirical risk minimization on real-world medical imaging data, in line with prior work on general imaging datasets. However, a subset of realistic induced-shift scenarios in clinical time series data do exhibit limited performance gains. We characterize these scenarios in detail, and recommend best practices for domain generalization in the clinical setting.

cs.LG↗
arXiv2020

CheXclusion: Fairness gaps in deep chest X-ray classifiers

Machine learning systems have received much attention recently for their ability to achieve expert-level performance on clinical tasks, particularly in medical imaging. Here, we examine the extent to which state-of-the-art deep learning classifiers trained to yield diagnostic labels from X-ray images are biased with respect to protected attributes. We train convolution neural networks to predict 14 diagnostic labels in 3 prominent public chest X-ray datasets: MIMIC-CXR, Chest-Xray8, CheXpert, as well as a multi-site aggregation of all those datasets. We evaluate the TPR disparity -- the difference in true positive rates (TPR) -- among different protected attributes such as patient sex, age, race, and insurance type as a proxy for socioeconomic status. We demonstrate that TPR disparities exist in the state-of-the-art classifiers in all datasets, for all clinical tasks, and all subgroups. A multi-source dataset corresponds to the smallest disparities, suggesting one way to reduce bias. We find that TPR disparities are not significantly correlated with a subgroup's proportional disease burden. As clinical models move from papers to products, we encourage clinical decision makers to carefully audit for algorithmic disparities prior to deployment. Our code can be found at, https://github.com/LalehSeyyed/CheXclusion

cs.CV↗
arXiv2020

Evaluating Knowledge Transfer in Neural Network for Medical Images

Deep learning and knowledge transfer techniques have permeated the field of medical imaging and are considered as key approaches for revolutionizing diagnostic imaging practices. However, there are still challenges for the successful integration of deep learning into medical imaging tasks due to a lack of large annotated imaging data. To address this issue, we propose a teacher-student learning framework to transfer knowledge from a carefully pre-trained convolutional neural network (CNN) teacher to a student CNN. In this study, we explore the performance of knowledge transfer in the medical imaging setting. We investigate the proposed network's performance when the student network is trained on a small dataset (target dataset) as well as when teacher's and student's domains are distinct. The performances of the CNN models are evaluated on three medical imaging datasets including Diabetic Retinopathy, CheXpert, and ChestX-ray8. Our results indicate that the teacher-student learning framework outperforms transfer learning for small imaging datasets. Particularly, the teacher-student learning framework improves the area under the ROC Curve (AUC) of the CNN model on a small sample of CheXpert (n=5k) by 4% and on ChestX-ray8 (n=5.6k) by 9%. In addition to small training data size, we also demonstrate a clear advantage of the teacher-student learning framework in the medical imaging setting compared to transfer learning. We observe that the teacher-student network holds a great promise not only to improve the performance of diagnosis but also to reduce overfitting when the dataset is small.

eess.IV↗