Xueru Zhang

Yaxuan Wang, Zhongteng Cai, Yujia Bao et al.

The rapid advancement of large language models (LLMs) has led to growing interest in using synthetic data to train future models. However, this creates a self-consuming retraining loop, where models are trained on their own outputs and may cause performance drops and induce emerging biases. In real-world applications, previously deployed LLMs may influence the data they generate, leading to a dynamic system driven by user feedback. For example, if a model continues to underserve users from a group, less query data will be collected from this particular demographic of users. In this study, we introduce the concept of \textbf{S}elf-\textbf{C}onsuming \textbf{P}erformative \textbf{L}oop (\textbf{SCPL}) and investigate the role of synthetic data in shaping bias during these dynamic iterative training processes under controlled performative feedback. This controlled setting is motivated by the inaccessibility of real-world user preference data from dynamic production systems, and enables us to isolate and analyze feedback-driven bias evolution in a principled manner. We focus on two types of loops, including the typical retraining setting and the incremental fine-tuning setting, which is largely underexplored. Through experiments on three real-world tasks, we find that the performative loop increases preference bias and decreases disparate bias. We design a reward-based rejection sampling strategy to mitigate the bias, moving towards more trustworthy self-improving systems.

Tian Xie, Xuwei Tan, Xueru Zhang

This paper studies algorithmic decision-making under human's strategic behavior, where a decision maker uses an algorithm to make decisions about human agents, and the latter with information about the algorithm may exert effort strategically and improve to receive favorable decisions. Unlike prior works that assume agents benefit from their efforts immediately, we consider realistic scenarios where the impacts of these efforts are persistent and agents benefit from efforts by making improvements gradually. We first develop a dynamic model to characterize persistent improvements and based on this construct a Stackelberg game to model the interplay between agents and the decision-maker. We analytically characterize the equilibrium strategies and identify conditions under which agents have incentives to improve. With the dynamics, we then study how the decision-maker can design an optimal policy to incentivize the largest improvements inside the agent population. We also extend the model to settings where 1) agents may be dishonest and game the algorithm into making favorable but erroneous decisions; 2) honest efforts are forgettable and not sufficient to guarantee persistent improvements. With the extended models, we further examine conditions under which agents prefer honest efforts over dishonest behavior and the impacts of forgettable efforts.

Tian Xie, Zhiqun Zuo, Mohammad Mahdi Khalili et al.

Machine learning systems have been widely used to make decisions about individuals who may behave strategically to receive favorable outcomes, e.g., they may genuinely improve the true labels or manipulate observable features directly to game the system without changing labels. Although both behaviors have been studied (often as two separate problems) in the literature, most works assume individuals can (i) perfectly foresee the outcomes of their behaviors when they best respond; (ii) change their features arbitrarily as long as it is affordable, and the costs they need to pay are deterministic functions of feature changes. In this paper, we consider a different setting and focus on imitative strategic behaviors with unforeseeable outcomes, i.e., individuals manipulate/improve by imitating the features of those with positive labels, but the induced feature changes are unforeseeable. We first propose a Stackelberg game to model the interplay between individuals and the decision-maker, under which we examine how the decision-maker's ability to anticipate individual behavior affects its objective function and the individual's best response. We show that the objective difference between the two can be decomposed into three interpretable terms, with each representing the decision-maker's preference for a certain behavior. By exploring the roles of each term, we theoretically illustrate how a decision-maker with adjusted preferences may simultaneously disincentivize manipulation, incentivize improvement, and promote fairness. Such theoretical results provide a guideline for decision-makers to inform better and socially responsible decisions in practice.

Tian Xie, Xueru Zhang

As machine learning (ML) models are increasingly used in social domains to make consequential decisions about humans, they often have the power to reshape data distributions. Humans, as strategic agents, continuously adapt their behaviors in response to the learning system. As populations change dynamically, ML systems may need frequent updates to ensure high performance. However, acquiring high-quality human-annotated samples can be highly challenging and even infeasible in social domains. A common practice to address this issue is using the model itself to annotate unlabeled data samples. This paper investigates the long-term impacts when ML models are retrained with model-annotated samples when they incorporate human strategic responses. We first formalize the interactions between strategic agents and the model and then analyze how they evolve under such dynamic interactions. We find that agents are increasingly likely to receive positive decisions as the model gets retrained, whereas the proportion of agents with positive labels may decrease over time. We thus propose a refined retraining process to stabilize the dynamics. Last, we examine how algorithmic fairness can be affected by these retraining processes and find that enforcing common fairness constraints at every round may not benefit the disadvantaged group in the long run. Experiments on (semi-)synthetic and real data validate the theoretical findings.

Tian Xie, Pavan Rauch, Xueru Zhang

When ML algorithms are deployed to automate human-related decisions, human agents may learn the underlying decision policies and adapt their behavior. Strategic Classification (SC) has emerged as a framework for studying this interaction between agents and decision-makers to design more trustworthy ML systems. Prior theoretical models in SC assume that agents are perfectly or approximately rational and respond to decision policies by optimizing their utility. However, the growing prevalence of LLMs raises the possibility that real-world agents may instead rely on these tools for strategic advice. This shift prompts two questions: (i) Can LLMs generate effective and socially responsible strategies in SC settings? (ii) Can existing SC theoretical models accurately capture agent behavior when agents follow LLM-generated advice? To investigate these questions, we examine five critical SC scenarios: hiring, loan applications, school admissions, personal income, and public assistance programs. We simulate agents with diverse profiles who interact with three commercial LLMs (GPT-4o, GPT-4.1, and GPT-5), following their suggestions on effort allocations on features. We compare the resulting agent behaviors with the best responses in existing SC models. Our findings show that: (i) Even without access to the decision policy, LLMs can generate effective strategies that improve both agents' scores and qualification; (ii) At the population level, LLM-guided effort allocation strategies yield similar or even higher score improvements, qualification rates, and fairness metrics as those predicted by the SC theoretical model, suggesting that the theoretical model may still serve as a reasonable proxy for LLM-influenced behavior; and (iii) At the individual level, LLMs tend to produce more diverse and balanced effort allocations than theoretical models.

Xuwei Tan, Ziyu Hu, Xueru Zhang

Machine learning models trained on real-world data often inherit and amplify biases against certain social groups, raising urgent concerns about their deployment at scale. While numerous bias mitigation methods have been proposed, comparing the effectiveness of bias mitigation methods remains difficult due to heterogeneous datasets, inconsistent fairness metrics, isolated evaluation of vision versus multi-modal models, and insufficient hyperparameter tuning that undermines fair comparisons. We introduce NH-Fair, a unified benchmark for fairness without harm that spans both vision models and large vision-language models (LVLMs) under standardized data, metrics, and training protocols, covering supervised and zero-shot regimes. Our key contributions are: (1) a systematic ERM tuning study that identifies training choices with large influence on both utility and disparities, yielding empirically grounded guidelines to help practitioners reduce expensive hyperparameter tuning space in achieving strong fairness and accuracy; (2) evidence that many debiasing methods do not reliably outperform a well-tuned ERM baseline, whereas a composite data-augmentation method consistently delivers parity gains without sacrificing utility, emerging as a promising practical strategy. (3) an analysis showing that while LVLMs achieve higher average accuracy, they still exhibit subgroup disparities, and gains from scaling are typically smaller than those from architectural or training-protocol choices. NH-Fair provides a reproducible, tuning-aware pipeline for rigorous, harm-aware fairness evaluation.

Zhongteng Cai, Mohammad Mahdi Khalili, Xueru Zhang

As machine learning (ML) algorithms are increasingly used in social domains to make predictions about humans, there is a growing concern that these algorithms may exhibit biases against certain social groups. Numerous notions of fairness have been proposed in the literature to measure the unfairness of ML. Among them, one class that receives the most attention is \textit{parity-based}, i.e., achieving fairness by equalizing treatment or outcomes for different social groups. However, achieving parity-based fairness often comes at the cost of lowering model accuracy and is undesirable for many high-stakes domains like healthcare. To avoid inferior accuracy, a line of research focuses on \textit{preference-based} fairness, under which any group of individuals would experience the highest accuracy and collectively prefer the ML outcomes assigned to them if they were given the choice between various sets of outcomes. However, these works assume individual demographic information is known and fully accessible during training. In this paper, we relax this requirement and propose a novel \textit{demographic-agnostic fairness without harm (DAFH)} optimization algorithm, which jointly learns a group classifier that partitions the population into multiple groups and a set of decoupled classifiers associated with these groups. Theoretically, we conduct sample complexity analysis and show that our method can outperform the baselines when demographic information is known and used to train decoupled classifiers. Experiments on both synthetic and real data validate the proposed method.

Tian Xie, Tongxin Yin, Vaishakh Keshava et al. · cmu

While large language models (LLMs) already play significant roles in society, research has shown that LLMs still generate content including social bias against certain sensitive groups. While existing benchmarks have effectively identified social biases in LLMs, a critical gap remains in our understanding of the underlying reasoning that leads to these biased outputs. This paper goes one step further to evaluate the causal reasoning process of LLMs when they answer questions eliciting social biases. We first propose a novel conceptual framework to classify the causal reasoning produced by LLMs. Next, we use LLMs to synthesize $1788$ questions covering $8$ sensitive attributes and manually validate them. The questions can test different kinds of causal reasoning by letting LLMs disclose their reasoning process with causal graphs. We then test 4 state-of-the-art LLMs. All models answer the majority of questions with biased causal reasoning, resulting in a total of $4135$ biased causal graphs. Meanwhile, we discover $3$ strategies for LLMs to avoid biased causal reasoning by analyzing the "bias-free" cases. Finally, we reveal that LLMs are also prone to "mistaken-biased" causal reasoning, where they first confuse correlation with causality to infer specific sensitive group names and then incorporate biased causal reasoning.