Machine Learning (ML) algorithms are increasingly used in our daily lives, yet often exhibit discrimination against protected groups. In this talk, I discuss the growing concern of bias in ML and overview existing approaches to address fairness issues. Then, I present three novel approaches developed by my research group. The first leverages generative AI to eliminate biases in training datasets, the second tackles non-convex problems arise in fair learning, and the third introduces a matrix decomposition-based post-processing approach to identify and eliminate unfair model components.

This paper presents a post-processing algorithm for training fair neural network regression models that satisfy statistical parity, utilizing an explainable singular value decomposition (SVD) of the weight matrix. We propose a linear transformation of the weight matrix, whereby the singular values derived from the SVD of the transformed matrix directly correspond to the differences in the first and second moments of the output distributions across two groups. Consequently, we can convert the fairness constraints into constraints on the singular values. We analytically solve the problem of finding the optimal weights under these constraints. Experimental validation on various datasets demonstrates that our method achieves a similar or superior fairness-accuracy trade-off compared to the baselines without using the sensitive attribute at the inference time.

Strategic classification, where individuals modify their features to influence machine learning (ML) decisions, presents critical fairness challenges. While group fairness in this setting has been widely studied, individual fairness remains underexplored. We analyze threshold-based classifiers and prove that deterministic thresholds violate individual fairness. Then, we investigate the possibility of using a randomized classifier to achieve individual fairness. We introduce conditions under which a randomized classifier ensures individual fairness and leverage these conditions to find an optimal and individually fair randomized classifier through a linear programming problem. Additionally, we demonstrate that our approach can be extended to group fairness notions. Experiments on real-world datasets confirm that our method effectively mitigates unfairness and improves the fairness-accuracy trade-off.

In several real-world applications (e.g., online advertising, item recommendations, etc.) it may not be possible to release and share the real dataset due to privacy concerns. As a result, synthetic data generation (SDG) has emerged as a promising solution for data sharing. While the main goal of private SDG is to create a dataset that preserves the privacy of individuals contributing to the dataset, the use of synthetic data also creates an opportunity to improve fairness. Since there often exist historical biases in the datasets, using the original real data for training can lead to an unfair model. Using synthetic data, we can attempt to remove such biases from the dataset before releasing the data. In this work, we formalize the definition of fairness in synthetic data generation and provide a general framework to achieve fairness. Then we consider two notions of counterfactual fairness and information filtering fairness and show how our framework can be used for these definitions.

Differential privacy mechanisms such as the Gaussian or Laplace mechanism have been widely used in data analytics for preserving individual privacy. However, they are mostly designed for continuous outputs and are unsuitable for scenarios where discrete values are necessary. Although various quantization mechanisms were proposed recently to generate discrete outputs under differential privacy, the outcomes are either biased or have an inferior accuracy-privacy trade-off. In this paper, we propose a family of quantization mechanisms that is unbiased and differentially private. It has a high degree of freedom and we show that some existing mechanisms can be considered as special cases of ours. To find the optimal mechanism, we formulate a linear optimization that can be solved efficiently using linear programming tools. Experiments show that our proposed mechanism can attain a better privacy-accuracy trade-off compared to baselines.

One approach for interpreting black-box machine learning models is to find a global approximation of the model using simple interpretable functions, which is called a metamodel (a model of the model). Approximating the black-box with a metamodel can be used to 1) estimate instance-wise feature importance; 2) understand the functional form of the model; 3) analyze feature interactions. In this work, we propose a new method for finding interpretable metamodels. Our approach utilizes Kolmogorov superposition theorem, which expresses multivariate functions as a composition of univariate functions (our primitive parameterized functions). This composition can be represented in the form of a tree. Inspired by symbolic regression, we use a modified form of genetic programming to search over different tree configurations. Gradient descent (GD) is used to optimize the parameters of a given configuration. Our method is a novel memetic algorithm that uses GD not only for training numerical constants but also for the training of building blocks. Using several experiments, we show that our method outperforms recent metamodeling approaches suggested for interpreting black-boxes.

Supervised learning models have been used in various domains such as lending, college admission, face recognition, natural language processing, etc. However, they may inherit pre-existing biases from training data and exhibit discrimination against protected social groups. Various fairness notions have been proposed to address unfairness issues. In this work, we focus on Equalized Loss (EL), a fairness notion that requires the expected loss to be (approximately) equalized across different groups. Imposing EL on the learning process leads to a non-convex optimization problem even if the loss function is convex, and the existing fair learning algorithms cannot properly be adopted to find the fair predictor under the EL constraint. This paper introduces an algorithm that can leverage off-the-shelf convex programming tools (e.g., CVXPY (Diamond and Boyd, 2016; Agrawal et al., 2018)) to efficiently find the global optimum of this non-convex optimization. In particular, we propose the ELminimizer algorithm, which finds the optimal fair predictor under EL by reducing the non-convex optimization to a sequence of convex optimization problems. We theoretically prove that our algorithm finds the global optimal solution under certain conditions. Then, we support our theoretical results through several empirical studies

The use of machine learning models in high-stake applications (e.g., healthcare, lending, college admission) has raised growing concerns due to potential biases against protected social groups. Various fairness notions and methods have been proposed to mitigate such biases. In this work, we focus on Counterfactual Fairness (CF), a fairness notion that is dependent on an underlying causal graph and first proposed by Kusner $\textit{et al.}$; it requires that the outcome an individual perceives is the same in the real world as it would be in a "counterfactual" world, in which the individual belongs to another social group. Learning fair models satisfying CF can be challenging. It was shown in (Kusner $\textit{et al.}$) that a sufficient condition for satisfying CF is to $\textbf{not}$ use features that are descendants of sensitive attributes in the causal graph. This implies a simple method that learns CF models only using non-descendants of sensitive attributes while eliminating all descendants. Although several subsequent works proposed methods that use all features for training CF models, there is no theoretical guarantee that they can satisfy CF. In contrast, this work proposes a new algorithm that trains models using all the available features. We theoretically and empirically show that models trained with this method can satisfy CF.

Many data analytics applications rely on temporal data, generated (and possibly acquired) sequentially for online analysis. How to release this type of data in a privacy-preserving manner is of great interest and more challenging than releasing one-time, static data. Because of the (potentially strong) temporal correlation within the data sequence, the overall privacy loss can accumulate significantly over time; an attacker with statistical knowledge of the correlation can be particularly hard to defend against. An idea that has been explored in the literature to mitigate this problem is to factor this correlation into the perturbation/noise mechanism. Existing work, however, either focuses on the offline setting (where perturbation is designed and introduced after the entire sequence has become available), or requires a priori information on the correlation in generating perturbation. In this study we propose an approach where the correlation is learned as the sequence is generated, and is used for estimating future data in the sequence. This estimate then drives the generation of the noisy released data. This method allows us to design better perturbation and is suitable for real-time operations. Using the notion of differential privacy, we show this approach achieves high accuracy with lower privacy loss compared to existing methods.

A common assumption in semi-supervised learning is that the labeled, unlabeled, and test data are drawn from the same distribution. However, this assumption is not satisfied in many applications. In many scenarios, the data is collected sequentially (e.g., healthcare) and the distribution of the data may change over time often exhibiting so-called covariate shifts. In this paper, we propose an approach for semi-supervised learning algorithms that is capable of addressing this issue. Our framework also recovers some popular methods, including entropy minimization and pseudo-labeling. We provide new information-theoretical based generalization error upper bounds inspired by our novel framework. Our bounds are applicable to both general semi-supervised learning and the covariate-shift scenario. Finally, we show numerically that our method outperforms previous approaches proposed for semi-supervised learning under the covariate shift.

Although machine learning (ML) algorithms are widely used to make decisions about individuals in various domains, concerns have arisen that (1) these algorithms are vulnerable to strategic manipulation and "gaming the algorithm"; and (2) ML decisions may exhibit bias against certain social groups. Existing works have largely examined these as two separate issues, e.g., by focusing on building ML algorithms robust to strategic manipulation, or on training a fair ML algorithm. In this study, we set out to understand the impact they each have on the other, and examine how to characterize fair policies in the presence of strategic behavior. The strategic interaction between a decision maker and individuals (as decision takers) is modeled as a two-stage (Stackelberg) game; when designing an algorithm, the former anticipates the latter may manipulate their features in order to receive more favorable decisions. We analytically characterize the equilibrium strategies of both, and examine how the algorithms and their resulting fairness properties are affected when the decision maker is strategic (anticipates manipulation), as well as the impact of fairness interventions on equilibrium strategies. In particular, we identify conditions under which anticipation of strategic behavior may mitigate/exacerbate unfairness, and conditions under which fairness interventions can serve as (dis)incentives for strategic manipulation.

We study the design of a class of incentive mechanisms that can effectively prevent cheating in a strategic classification and regression problem. A conventional strategic classification or regression problem is modeled as a Stackelberg game, or a principal-agent problem between the designer of a classifier (the principal) and individuals subject to the classifier's decisions (the agents), potentially from different demographic groups. The former benefits from the accuracy of its decisions, whereas the latter may have an incentive to game the algorithm into making favorable but erroneous decisions. While prior works tend to focus on how to design an algorithm to be more robust to such strategic maneuvering, this study focuses on an alternative, which is to design incentive mechanisms to shape the utilities of the agents and induce effort that genuinely improves their skills, which in turn benefits both parties in the Stackelberg game. Specifically, the principal and the mechanism provider (which could also be the principal itself) move together in the first stage, publishing and committing to a classifier and an incentive mechanism. The agents are (simultaneous) second movers and best respond to the published classifier and incentive mechanism. When an agent's strategic action merely changes its observable features, it hurts the performance of the algorithm. However, if the action leads to improvement in the agent's true label, it not only helps the agent achieve better decision outcomes, but also preserves the performance of the algorithm. We study how a subsidy mechanism can induce improvement actions, positively impact a number of social well-being metrics, such as the overall skill levels of the agents (efficiency) and positive or true positive rate differences between different demographic groups (fairness).

Supervised learning models have been increasingly used for making decisions about individuals in applications such as hiring, lending, and college admission. These models may inherit pre-existing biases from training datasets and discriminate against protected attributes (e.g., race or gender). In addition to unfairness, privacy concerns also arise when the use of models reveals sensitive personal information. Among various privacy notions, differential privacy has become popular in recent years. In this work, we study the possibility of using a differentially private exponential mechanism as a post-processing step to improve both fairness and privacy of supervised learning models. Unlike many existing works, we consider a scenario where a supervised model is used to select a limited number of applicants as the number of available positions is limited. This assumption is well-suited for various scenarios, such as job application and college admission. We use ``equal opportunity'' as the fairness notion and show that the exponential mechanisms can make the decision-making process perfectly fair. Moreover, the experiments on real-world datasets show that the exponential mechanism can improve both privacy and fairness, with a slight decrease in accuracy compared to the model without post-processing.

We consider a selection problem where sequentially arrived applicants apply for a limited number of positions/jobs. At each time step, a decision maker accepts or rejects the given applicant using a pre-trained supervised learning model until all the vacant positions are filled. In this paper, we discuss whether the fairness notions (e.g., equal opportunity, statistical parity, etc.) that are commonly used in classification problems are suitable for the sequential selection problems. In particular, we show that even with a pre-trained model that satisfies the common fairness notions, the selection outcomes may still be biased against certain demographic groups. This observation implies that the fairness notions used in classification problems are not suitable for a selection problem where the applicants compete for a limited number of positions. We introduce a new fairness notion, ``Equal Selection (ES),'' suitable for sequential selection problems and propose a post-processing approach to satisfy the ES fairness notion. We also consider a setting where the applicants have privacy concerns, and the decision maker only has access to the noisy version of sensitive attributes. In this setting, we can show that the \textit{perfect} ES fairness can still be attained under certain conditions.

Alternating direction method of multiplier (ADMM) is a powerful method to solve decentralized convex optimization problems. In distributed settings, each node performs computation with its local data and the local results are exchanged among neighboring nodes in an iterative fashion. During this iterative process the leakage of data privacy arises and can accumulate significantly over many iterations, making it difficult to balance the privacy-accuracy tradeoff. We propose Recycled ADMM (R-ADMM), where a linear approximation is applied to every even iteration, its solution directly calculated using only results from the previous, odd iteration. It turns out that under such a scheme, half of the updates incur no privacy loss and require much less computation compared to the conventional ADMM. Moreover, R-ADMM can be further modified (MR-ADMM) such that each node independently determines its own penalty parameter over iterations. We obtain a sufficient condition for the convergence of both algorithms and provide the privacy analysis based on objective perturbation. It can be shown that the privacy-accuracy tradeoff can be improved significantly compared with conventional ADMM.

Recognizing the factors that cause stress is a crucial step toward early detection of stressors. In this regard, several studies make an effort to recognize individuals' stress using an Electroencephalogram (EEG). However, current EEG-based stress recognition frameworks have several drawbacks. First, they are mostly designed to recognize individuals' stress only in a controlled laboratory environment. Second, they do not take into account the changes in the EEG signals of different subjects under the same stressors. Third, most of the current stress recognition algorithms occur in an offline setting. To address these issues, this study proposes an EEG-based stress recognition framework that takes into account each subject's brainwave patterns to train the stress recognition classifier and continuously update its classifier based on new input signals in near real-time. The proposed framework first removes EEG signal artifacts, then extracts a broad range of EEG signal features, and finally applies different online multitask learning (OMTL) algorithms to recognize individuals' stress in near real time. The proposed framework was applied on the EEG collected in two environments-first on the EEG collected in a controlled lab environment using a wired-EEG and second on the EEG collected at in the field using a wearable EEG device. The OMTL-VonNeuman method resulted in the best prediction accuracy on both datasets (71.14% on the first dataset and 77.61% on second) among all tested algorithms. The proposed stress recognition framework continuously updates its classifier and therefore contributes to stress recognition for new stressful situations that are beyond the range of predefined stressful conditions in near real time both in a controlled lab environment and at real job sites.

Alternating direction method of multiplier (ADMM) is a popular method used to design distributed versions of a machine learning algorithm, whereby local computations are performed on local data with the output exchanged among neighbors in an iterative fashion. During this iterative process the leakage of data privacy arises. A differentially private ADMM was proposed in prior work (Zhang & Zhu, 2017) where only the privacy loss of a single node during one iteration was bounded, a method that makes it difficult to balance the tradeoff between the utility attained through distributed computation and privacy guarantees when considering the total privacy loss of all nodes over the entire iterative process. We propose a perturbation method for ADMM where the perturbed term is correlated with the penalty parameters; this is shown to improve the utility and privacy simultaneously. The method is based on a modified ADMM where each node independently determines its own penalty parameter in every iteration and decouples it from the dual updating step size. The condition for convergence of the modified ADMM and the lower bound on the convergence rate are also derived.

Cyber insurance is a viable method for cyber risk transfer. However, it has been shown that depending on the features of the underlying environment, it may or may not improve the state of network security. In this paper, we consider a single profit-maximizing insurer (principal) with voluntarily participating insureds/clients (agents). We are particularly interested in two distinct features of cybersecurity and their impact on the contract design problem. The first is the interdependent nature of cybersecurity, whereby one entity's state of security depends not only on its own investment and effort, but also the efforts of others' in the same eco-system (i.e., externalities). The second is the fact that recent advances in Internet measurement combined with machine learning techniques now allow us to perform accurate quantitative assessments of security posture at a firm level. This can be used as a tool to perform an initial security audit, or pre-screening, of a prospective client to better enable premium discrimination and the design of customized policies. We show that security interdependency leads to a “profit opportunity” for the insurer, created by the inefficient effort levels exerted by interdependent agents who do not account for the risk externalities when insurance is not available; this is in addition to risk transfer that an insurer typically profits from. Security pre-screening then allows the insurer to take advantage of this additional profit opportunity by designing the appropriate contracts which incentivize agents to increase their effort levels, allowing the insurer to “sell commitment” to interdependent agents, in addition to insuring their risks. We identify conditions under which this type of contract leads to not only increased profit for the principal, but also an improved state of network security.