The rapid integration of Large Language Models (LLMs) into various domains raises concerns about societal inequalities and information bias. This study examines biases in LLMs related to background, gender, and age, with a focus on their impact on decision-making and summarization tasks. Additionally, the research examines the cross-lingual propagation of these biases and evaluates the effectiveness of prompt-instructed mitigation strategies. Using an adapted version of the dataset by Tamkin et al. (2023) translated into Dutch, we created 151,200 unique prompts for the decision task and 176,400 for the summarisation task. Various demographic variables, instructions, salience levels, and languages were tested on GPT-3.5 and GPT-4o. Our analysis revealed that both models were significantly biased during decision-making, favouring female gender, younger ages, and certain backgrounds such as the African-American background. In contrast, the summarisation task showed minimal evidence of bias, though significant age-related differences emerged for GPT-3.5 in English. Cross-lingual analysis showed that bias patterns were broadly similar between English and Dutch, though notable differences were observed across specific demographic categories. The newly proposed mitigation instructions, while unable to eliminate biases completely, demonstrated potential in reducing them. The most effective instruction achieved a 27\% mean reduction in the gap between the most and least favorable demographics. Notably, contrary to GPT-3.5, GPT-4o displayed reduced biases for all prompts in English, indicating the specific potential for prompt-based mitigation within newer models. This research underscores the importance of cautious adoption of LLMs and context-specific bias testing, highlighting the need for continued development of effective mitigation strategies to ensure responsible deployment of AI.

Large language models (LLMs) exhibit cognitive biases -- systematic tendencies of irrational decision-making, similar to those seen in humans. Prior work has found that these biases vary across models and can be amplified by instruction tuning. However, it remains unclear if these differences in biases stem from pretraining, finetuning, or even random noise due to training stochasticity. We propose a two-step causal experimental approach to disentangle these factors. First, we finetune models multiple times using different random seeds to study how training randomness affects over $30$ cognitive biases. Second, we introduce \emph{cross-tuning} -- swapping instruction datasets between models to isolate bias sources. This swap uses datasets that led to different bias patterns, directly testing whether biases are dataset-dependent. Our findings reveal that while training randomness introduces some variability, biases are mainly shaped by pretraining: models with the same pretrained backbone exhibit more similar bias patterns than those sharing only finetuning data. These insights suggest that understanding biases in finetuned models requires considering their pretraining origins beyond finetuning effects. This perspective can guide future efforts to develop principled strategies for evaluating and mitigating bias in LLMs.

Most machine learning classifiers only concern classification accuracy, while certain applications (such as medical diagnosis, meteorological forecasting, and computation advertising) require the model to predict the true probability, known as a calibrated estimate. In previous work, researchers have developed several calibration methods to post-process the outputs of a predictor to obtain calibrated values, such as binning and scaling methods. Compared with scaling, binning methods are shown to have distribution-free theoretical guarantees, which motivates us to prefer binning methods for calibration. However, we notice that existing binning methods have several drawbacks: (a) the binning scheme only considers the original prediction values, thus limiting the calibration performance; and (b) the binning approach is non-individual, mapping multiple samples in a bin to the same value, and thus is not suitable for order-sensitive applications. In this paper, we propose a feature-aware binning framework, called Multiple Boosting Calibration Trees (MBCT), along with a multi-view calibration loss to tackle the above issues. Our MBCT optimizes the binning scheme by the tree structures of features, and adopts a linear function in a tree node to achieve individual calibration. Our MBCT is non-monotonic, and has the potential to improve order accuracy, due to its learnable binning scheme and the individual calibration. We conduct comprehensive experiments on three datasets in different fields. Results show that our method outperforms all competing models in terms of both calibration error and order accuracy. We also conduct simulation experiments, justifying that the proposed multi-view calibration loss is a better metric in modeling calibration error.

Large language models (LLMs) have shown impressive potential in helping with numerous medical challenges. Deploying LLMs in high-stakes applications such as medicine, however, brings in many concerns. One major area of concern relates to biased behaviors of LLMs in medical applications, leading to unfair treatment of individuals. To pave the way for the responsible and impactful deployment of Med LLMs, rigorous evaluation is a key prerequisite. Due to the huge complexity and variability of different medical scenarios, existing work in this domain has primarily relied on using manually crafted datasets for bias evaluation. In this study, we present a new method to scale up such bias evaluations by automatically generating test cases based on rigorous medical evidence. We specifically target the challenges of a) domain-specificity of bias characterization, b) hallucinating while generating the test cases, and c) various dependencies between the health outcomes and sensitive attributes. To that end, we offer new methods to address these challenges integrated with our generative pipeline, using medical knowledge graphs, medical ontologies, and customized general LLM evaluation frameworks in our method. Through a series of extensive experiments, we show that the test cases generated by our proposed method can effectively reveal bias patterns in Med LLMs at larger and more flexible scales than human-crafted datasets. We publish a large bias evaluation dataset using our pipeline, which is dedicated to a few medical case studies. A live demo of our application for vignette generation is available at https://vignette.streamlit.app. Our code is also available at https://github.com/healthylaife/autofair.

Although achieving promising performance, recent analyses show that current generative large language models (LLMs) may still capture dataset biases and utilize them for generation, leading to poor generalizability and harmfulness of LLMs. However, due to the diversity of dataset biases and the over-optimization problem, previous prior-knowledge-based debiasing methods and fine-tuning-based debiasing methods may not be suitable for current LLMs. To address this issue, we explore combining active learning with the causal mechanisms and propose a casual-guided active learning (CAL) framework, which utilizes LLMs itself to automatically and autonomously identify informative biased samples and induce the bias patterns. Then a cost-effective and efficient in-context learning based method is employed to prevent LLMs from utilizing dataset biases during generation. Experimental results show that CAL can effectively recognize typical biased instances and induce various bias patterns for debiasing LLMs.

Large Language Models (LLMs) have emerged as promising solutions for a variety of medical and clinical decision support applications. However, LLMs are often subject to different types of biases, which can lead to unfair treatment of individuals, worsening health disparities, and reducing trust in AI-augmented medical tools. Aiming to address this important issue, in this study, we present a new model alignment approach for aligning LLMs using a preference optimization method within a knowledge distillation framework. Prior to presenting our proposed method, we first use an evaluation framework to conduct a comprehensive (largest to our knowledge) empirical evaluation to reveal the type and nature of existing biases in LLMs used for medical applications. We then offer a bias mitigation technique to reduce the unfair patterns in LLM outputs across different subgroups identified by the protected attributes. We show that our mitigation method is effective in significantly reducing observed biased patterns. Our code is publicly available at \url{https://github.com/healthylaife/FairAlignmentLLM}.

Large Language Models (LLMs) have emerged as powerful candidates to inform clinical decision-making processes. While these models play an increasingly prominent role in shaping the digital landscape, two growing concerns emerge in healthcare applications: 1) to what extent do LLMs exhibit social bias based on patients' protected attributes (like race), and 2) how do design choices (like architecture design and prompting strategies) influence the observed biases? To answer these questions rigorously, we evaluated eight popular LLMs across three question-answering (QA) datasets using clinical vignettes (patient descriptions) standardized for bias evaluations. We employ red-teaming strategies to analyze how demographics affect LLM outputs, comparing both general-purpose and clinically-trained models. Our extensive experiments reveal various disparities (some significant) across protected groups. We also observe several counter-intuitive patterns such as larger models not being necessarily less biased and fined-tuned models on medical data not being necessarily better than the general-purpose models. Furthermore, our study demonstrates the impact of prompt design on bias patterns and shows that specific phrasing can influence bias patterns and reflection-type approaches (like Chain of Thought) can reduce biased outcomes effectively. Consistent with prior studies, we call on additional evaluations, scrutiny, and enhancement of LLMs used in clinical decision support applications.

Large pre-trained language models have shown remarkable performance over the past few years. These models, however, sometimes learn superficial features from the dataset and cannot generalize to the distributions that are dissimilar to the training scenario. There have been several approaches proposed to reduce model's reliance on these bias features which can improve model robustness in the out-of-distribution setting. However, existing methods usually use a fixed low-capacity model to deal with various bias features, which ignore the learnability of those features. In this paper, we analyze a set of existing bias features and demonstrate there is no single model that works best for all the cases. We further show that by choosing an appropriate bias model, we can obtain a better robustness result than baselines with a more sophisticated model design.

Natural Language Sentence Matching (NLSM) has gained substantial attention from both academics and the industry, and rich public datasets contribute a lot to this process. However, biased datasets can also hurt the generalization performance of trained models and give untrustworthy evaluation results. For many NLSM datasets, the providers select some pairs of sentences into the datasets, and this sampling procedure can easily bring unintended pattern, i.e., selection bias. One example is the QuoraQP dataset, where some content-independent naive features are unreasonably predictive. Such features are the reflection of the selection bias and termed as the leakage features. In this paper, we investigate the problem of selection bias on six NLSM datasets and find that four out of them are significantly biased. We further propose a training and evaluation framework to alleviate the bias. Experimental results on QuoraQP suggest that the proposed framework can improve the generalization ability of trained models, and give more trustworthy evaluation results for real-world adoptions.