Toward Automatic Group Membership Annotation for Group Fairness Evaluation

With the increasing research attention on fairness in information retrieval systems, more and more fairness-aware algorithms have been proposed to ensure fairness for a sustainable and healthy retrieval ecosystem. However, as the most adopted measurement of fairness-aware algorithms, group fairness evaluation metrics, require group membership information that needs massive human annotations and is barely available for general information retrieval datasets. This data sparsity significantly impedes the development of fairness-aware information retrieval studies. Hence, a practical, scalable, low-cost group membership annotation method is needed to assist or replace human annotations. This study explored how to leverage language models to automatically annotate group membership for group fairness evaluations, focusing on annotation accuracy and its impact. Our experimental results show that BERT-based models outperformed state-of-the-art large language models, including GPT and Mistral, achieving promising annotation accuracy with minimal supervision in recent fair-ranking datasets. Our impact-oriented evaluations reveal that minimal annotation error will not degrade the effectiveness and robustness of group fairness evaluation. The proposed annotation method reduces tremendous human efforts and expands the frontier of fairness-aware studies to more datasets.

Learn to be Fair without Labels: a Distribution-based Learning Framework for Fair Ranking

Ranking algorithms as an essential component of retrieval systems have been constantly improved in previous studies, especially regarding relevance-based utilities. In recent years, more and more research attempts have been proposed regarding fairness in rankings due to increasing concerns about potential discrimination and the issue of echo chamber. These attempts include traditional score-based methods that allocate exposure resources to different groups using pre-defined scoring functions or selection strategies and learning-based methods that learn the scoring functions based on data samples. Learning-based models are more flexible and achieve better performance than traditional methods. However, most of the learning-based models were trained and tested on outdated datasets where fairness labels are barely available. State-of-art models utilize relevance-based utility scores as a substitute for the fairness labels to train their fairness-aware loss, where plugging in the substitution does not guarantee the minimum loss. This inconsistency challenges the model's accuracy and performance, especially when learning is achieved by gradient descent. Hence, we propose a distribution-based fair learning framework (DLF) that does not require labels by replacing the unavailable fairness labels with target fairness exposure distributions. Experimental studies on TREC fair ranking track dataset confirm that our proposed framework achieves better fairness performance while maintaining better control over the fairness-relevance trade-off than state-of-art fair ranking frameworks.

Behavior Alignment: A New Perspective of Evaluating LLM-based Conversational Recommendation Systems

Large Language Models (LLMs) have demonstrated great potential in Conversational Recommender Systems (CRS). However, the application of LLMs to CRS has exposed a notable discrepancy in behavior between LLM-based CRS and human recommenders: LLMs often appear inflexible and passive, frequently rushing to complete the recommendation task without sufficient inquiry.This behavior discrepancy can lead to decreased accuracy in recommendations and lower user satisfaction. Despite its importance, existing studies in CRS lack a study about how to measure such behavior discrepancy. To fill this gap, we propose Behavior Alignment, a new evaluation metric to measure how well the recommendation strategies made by a LLM-based CRS are consistent with human recommenders'. Our experiment results show that the new metric is better aligned with human preferences and can better differentiate how systems perform than existing evaluation metrics. As Behavior Alignment requires explicit and costly human annotations on the recommendation strategies, we also propose a classification-based method to implicitly measure the Behavior Alignment based on the responses. The evaluation results confirm the robustness of the method.