Haewon Jeong

Wael Alghamdi, Hsiang Hsu, Haewon Jeong et al.

We consider the problem of producing fair probabilistic classifiers for multi-class classification tasks. We formulate this problem in terms of "projecting" a pre-trained (and potentially unfair) classifier onto the set of models that satisfy target group-fairness requirements. The new, projected model is given by post-processing the outputs of the pre-trained classifier by a multiplicative factor. We provide a parallelizable iterative algorithm for computing the projected classifier and derive both sample complexity and convergence guarantees. Comprehensive numerical comparisons with state-of-the-art benchmarks demonstrate that our approach maintains competitive performance in terms of accuracy-fairness trade-off curves, while achieving favorable runtime on large datasets. We also evaluate our method at scale on an open dataset with multiple classes, multiple intersectional protected groups, and over 1M samples.

Sajani Vithana, Viveck R. Cadambe, Flavio P. Calmon et al.

Differentially private distributed mean estimation (DP-DME) is a fundamental building block in privacy-preserving federated learning, where a central server estimates the mean of $d$-dimensional vectors held by $n$ users while ensuring $(ε,δ)$-DP. Local differential privacy (LDP) and distributed DP with secure aggregation (SA) are the most common notions of DP used in DP-DME settings with an untrusted server. LDP provides strong resilience to dropouts, colluding users, and adversarial attacks, but suffers from poor utility. In contrast, SA-based DP-DME achieves an $O(n)$ utility gain over LDP in DME, but requires increased communication and computation overheads and complex multi-round protocols to handle dropouts and attacks. In this work, we present a generalized framework for DP-DME, that captures LDP and SA-based mechanisms as extreme cases. Our framework provides a foundation for developing and analyzing a variety of DP-DME protocols that leverage correlated privacy mechanisms across users. To this end, we propose CorDP-DME, a novel DP-DME mechanism based on the correlated Gaussian mechanism, that spans the gap between DME with LDP and distributed DP. We prove that CorDP-DME offers a favorable balance between utility and resilience to dropout and collusion. We provide an information-theoretic analysis of CorDP-DME, and derive theoretical guarantees for utility under any given privacy parameters and dropout/colluding user thresholds. Our results demonstrate that (anti) correlated Gaussian DP mechanisms can significantly improve utility in mean estimation tasks compared to LDP -- even in adversarial settings -- while maintaining better resilience to dropouts and attacks compared to distributed DP.

Viveck R. Cadambe, Ateet Devulapalli, Haewon Jeong et al.

We consider the problem of private distributed multi-party multiplication. It is well-established that Shamir secret-sharing coding strategies can enable perfect information-theoretic privacy in distributed computation via the celebrated algorithm of Ben Or, Goldwasser and Wigderson (the "BGW algorithm"). However, perfect privacy and accuracy require an honest majority, that is, $N \geq 2t+1$ compute nodes are required to ensure privacy against any $t$ colluding adversarial nodes. By allowing for some controlled amount of information leakage and approximate multiplication instead of exact multiplication, we study coding schemes for the setting where the number of honest nodes can be a minority, that is $N< 2t+1.$ We develop a tight characterization privacy-accuracy trade-off for cases where $N < 2t+1$ by measuring information leakage using {differential} privacy instead of perfect privacy, and using the mean squared error metric for accuracy. A novel technical aspect is an intricately layered noise distribution that merges ideas from differential privacy and Shamir secret-sharing at different layers.

Youngseok Yoon, Dainong Hu, Iain Weissburg et al.

Model collapse, the severe degradation of generative models when iteratively trained on their own outputs, has gained significant attention in recent years. This paper examines Chain of Diffusion, where a pretrained text-to-image diffusion model is finetuned on its own generated images. We demonstrate that severe image quality degradation was universal and identify CFG scale as the key factor impacting this model collapse. Drawing on an analogy between the Chain of Diffusion and biological evolution, we then introduce a novel theoretical analysis based on quantitative trait modeling from statistical genetics. Our theoretical analysis aligns with empirical observations of the generated images in the Chain of Diffusion. Finally, we propose Reusable Diffusion Finetuning (ReDiFine), a simple yet effective strategy inspired by genetic mutations. It operates robustly across various scenarios without requiring any hyperparameter tuning, making it a plug-and-play solution for reusable image generation.

Arjun Nichani, Hsiang Hsu, Chun-Fu et al.

Fairness and privacy are two vital pillars of trustworthy machine learning. Despite extensive research on these individual topics, the relationship between fairness and privacy has received significantly less attention. In this paper, we utilize the information-theoretic measure Chernoff Information to highlight the data-dependent nature of the relationship among the triad of fairness, privacy, and accuracy. We first define Noisy Chernoff Difference, a tool that allows us to analyze the relationship among the triad simultaneously. We then show that for synthetic data, this value behaves in 3 distinct ways (depending on the distribution of the data). We highlight the data distributions involved in these cases and explore their fairness and privacy implications. Additionally, we show that Noisy Chernoff Difference acts as a proxy for the steepness of the fairness-accuracy curves. Finally, we propose a method for estimating Chernoff Information on data from unknown distributions and utilize this framework to examine the triad dynamic on real datasets. This work builds towards a unified understanding of the fairness-privacy-accuracy relationship and highlights its data-dependent nature.

Iain Weissburg, Sathvika Anand, Sharon Levy et al.

With the increasing adoption of large language models (LLMs) in education, concerns about inherent biases in these models have gained prominence. We evaluate LLMs for bias in the personalized educational setting, specifically focusing on the models' roles as "teachers." We reveal significant biases in how models generate and select educational content tailored to different demographic groups, including race, ethnicity, sex, gender, disability status, income, and national origin. We introduce and apply two bias score metrics--Mean Absolute Bias (MAB) and Maximum Difference Bias (MDB)--to analyze 9 open and closed state-of-the-art LLMs. Our experiments, which utilize over 17,000 educational explanations across multiple difficulty levels and topics, uncover that models potentially harm student learning by both perpetuating harmful stereotypes and reversing them. We find that bias is similar for all frontier models, with the highest MAB along income levels while MDB is highest relative to both income and disability status. For both metrics, we find the lowest bias exists for sex/gender and race/ethnicity.

Tian Qiu, Arjun Nichani, Rasta Tadayontahmasebi et al.

Neural compression methods are gaining popularity due to their superior rate-distortion performance over traditional methods, even at extremely low bitrates below 0.1 bpp. As deep learning architectures, these models are prone to bias during the training process, potentially leading to unfair outcomes for individuals in different groups. In this paper, we present a general, structured, scalable framework for evaluating bias in neural image compression models. Using this framework, we investigate racial bias in neural compression algorithms by analyzing nine popular models and their variants. Through this investigation, we first demonstrate that traditional distortion metrics are ineffective in capturing bias in neural compression models. Next, we highlight that racial bias is present in all neural compression models and can be captured by examining facial phenotype degradation in image reconstructions. We then examine the relationship between bias and realism in the decoded images and demonstrate a trade-off across models. Finally, we show that utilizing a racially balanced training set can reduce bias but is not a sufficient bias mitigation strategy. We additionally show the bias can be attributed to compression model bias and classification model bias. We believe that this work is a first step towards evaluating and eliminating bias in neural image compression models.

Louisa Cornelis, Guillermo Bernárdez, Haewon Jeong et al.

Personalization in machine learning involves tailoring models to individual users by incorporating personal attributes such as demographic or medical data. While personalization can improve prediction accuracy, it may also amplify biases and reduce explainability. This work introduces a unified framework to evaluate the impact of personalization on both prediction accuracy and explanation quality across classification and regression tasks. We derive novel upper bounds for the number of personal attributes that can be used to reliably validate benefits of personalization. Our analysis uncovers key trade-offs. We show that regression models can potentially utilize more personal attributes than classification models. We also demonstrate that improvements in prediction accuracy due to personalization do not necessarily translate to enhanced explainability -- underpinning the importance to evaluate both metrics when personalizing machine learning models in critical settings such as healthcare. Validated with a real-world dataset, this framework offers practical guidance for balancing accuracy, fairness, and interpretability in personalized models.

Haewon Jeong, Hao Wang, Flavio P. Calmon

We investigate the fairness concerns of training a machine learning model using data with missing values. Even though there are a number of fairness intervention methods in the literature, most of them require a complete training set as input. In practice, data can have missing values, and data missing patterns can depend on group attributes (e.g. gender or race). Simply applying off-the-shelf fair learning algorithms to an imputed dataset may lead to an unfair model. In this paper, we first theoretically analyze different sources of discrimination risks when training with an imputed dataset. Then, we propose an integrated approach based on decision trees that does not require a separate process of imputation and learning. Instead, we train a tree with missing incorporated as attribute (MIA), which does not require explicit imputation, and we optimize a fairness-regularized objective function. We demonstrate that our approach outperforms existing fairness intervention methods applied to an imputed dataset, through several experiments on real-world datasets.