Session 2: Bias and Fairness 1

Date: Tuesday October 15, 12:00 PM – 13:15 PM (GMT+2)
Room: Petruzzelli Theater
Session Chair: Vito Walter Anelli

• RES 🕓5FairCRS: Towards User-oriented Fairness in Conversational Recommendation Systems
  by Qin Liu (Jinan University), Xuan Feng (Jinan University), Tianlong Gu (Jinan University) and Xiaoli Liu (Jinan University)

  Conversational Recommendation Systems (CRSs) enable recommender systems to explicitly acquire user preferences during multi-turn interactions, providing more accurate and personalized recommendations. However, the data imbalance in CRSs, due to inconsistent interaction history among users, may lead to disparate treatment for disadvantaged user groups. In this paper, we investigate the discriminate problems in CRS from the user’s perspective, called as user-oriented fairness. To reveal the unfairness problems of different user groups in CRS, we conduct extensive empirical analyses. To mitigate user unfairness, we propose a user-oriented fairness framework, named FairCRS, which is a model-agnostic framework. In particular, we develop a user-embedding reconstruction mechanism that enriches user embeddings by incorporating more interaction information, and design a user-oriented fairness strategy that optimizes the recommendation quality differences among user groups while alleviating unfairness. Extensive experimental results on English and Chinese datasets show that FairCRS outperforms state-of-the-art CRSs in terms of overall recommendation performance and user fairness.

  Full text in ACM Digital Library

• REPR 🕓10AMBAR: A dataset for Assessing Multiple Beyond-Accuracy Recommenders
  by Elizabeth Gómez (Universitat de Barcelona), David Contreras (Universidad Arturo Prat), Ludovico Boratto (University of Cagliari) and Maria Salamo (Universitat de Barcelona)

  Nowadays a recommendation model should exploit additional information from both the user and item perspectives, in addition to utilizing user-item interaction data. Datasets are central in offering the required information for evaluating new models or algorithms. Although there are many datasets in the literature with user and item properties, there are several issues not covered yet: (i) it is difficult to perform cross-analysis of properties at user and item level as they are not related in most cases; and (ii) on top of that, in many occasions datasets do not allow analysis at different granularity levels. In this paper, we propose a new dataset in the music domain, named AMBAR, that tackles the above-mentioned issues. Besides detailing in depth the structure of the new dataset, we also show its application in contexts (i.e., multi-objective, fair, and calibrated recommendations) where both the effectiveness and the beyond-accuracy perspectives of recommendation are assessed.

  Full text in ACM Digital Library

• REPR 🕓10Do Recommender Systems Promote Local Music? A Reproducibility Study Using Music Streaming Data
  by Kristina Matrosova (Geographie-Cités), Lilian Marey (Télécom Paris), Guillaume Salha-Galvan (Deezer Research), Thomas Louail (Geographie-Cités), Olivier Bodini (Université Sorbonne Paris Nord) and Manuel Moussallam (Deezer Research)

  This paper examines the influence of recommender systems on local music representation, discussing prior findings from an empirical study on the LFM-2b public dataset. This prior study argued that different recommender systems exhibit algorithmic biases shifting music consumption either towards or against local content. However, LFM-2b users do not reflect the diverse audience of music streaming services. To assess the robustness of this study’s conclusions, we conduct a comparative analysis using proprietary listening data from a global music streaming service, which we publicly release alongside this paper. We observe significant differences in local music consumption patterns between our dataset and LFM-2b, suggesting that caution should be exercised when drawing conclusions on local music based solely on LFM-2b. Moreover, we show that the algorithmic biases exhibited in the original work vary in our dataset, and that several unexplored model parameters can significantly influence these biases and affect the study’s conclusion on both datasets. Finally, we discuss the complexity of accurately labeling local music, emphasizing the risk of misleading conclusions due to unreliable, biased, or incomplete labels. To encourage further research and ensure reproducibility, we have publicly shared our dataset and code.

  Full text in ACM Digital Library

• REPR 🕓10Fair Augmentation for Graph Collaborative Filtering
  by Ludovico Boratto (University of Cagliari), Francesco Fabbri (Spotify), Gianni Fenu (University of Cagliari), Mirko Marras (University of Cagliari) and Giacomo Medda (University of Cagliari)

  Recent developments in recommendation have harnessed the collaborative power of graph neural networks (GNNs) in learning users’ preferences from user-item networks. Despite emerging regulations addressing fairness of automated systems, unfairness issues in graph collaborative filtering remain underexplored, especially from the consumer’s perspective. Despite numerous contributions on consumer unfairness, only a few of these works have delved into GNNs. A notable gap exists in the formalization of the latest mitigation algorithms, as well as in their effectiveness and reliability on cutting-edge models. This paper serves as a solid response to recent research highlighting unfairness issues in graph collaborative filtering by reproducing one of the latest mitigation methods. The reproduced technique adjusts the system fairness level by learning a fair graph augmentation. Under an experimental setup based on 11 GNNs, 5 non-GNN models, and 5 real-world networks across diverse domains, our investigation reveals that fair graph augmentation is consistently effective on high-utility models and large datasets. Experiments on the transferability of the fair augmented graph open new issues for future recommendation studies. Source code: https://github.com/jackmedda/FA4GCF.

  Full text in ACM Digital Library

• IND 🕓10Enhancing Recommendation Quality of the SASRec Model by Mitigating Popularity Bias
  by Venkata Harshit Koneru (ZDF), Xenija Neufeld (Accso – Accelerated Solutions GmbH), Sebastian Loth (ZDF) and Andreas Grün (ZDF)

  ZDF is a Public Service Media (PSM) broadcaster in Germany that uses recommender systems on its streaming service platform ZDFmediathek. One of the main use cases within the ZDFmediathek is Next Video, which is currently based on a Self-Attention based Sequential Recommendation model (SASRec). For this use case, we modified the loss function, the sampling method of negative items, and introduced the top-k negative sampling strategy and compared this to the vanilla SASRec model. We show that this not only reduces popularity bias, but also increases clicks and viewing volume compared to that of the vanilla version.

  Full text in ACM Digital Library

• RES 🕓15Putting Popularity Bias Mitigation to the Test: A User-Centric Evaluation in Music Recommenders
  by Robin Ungruh (Delft University of Technology), Karlijn Dinnissen (Utrecht University), Anja Volk (Utrecht University), Maria Soledad Pera (Delft University of Technology) and Hanna Hauptmann (Utrecht University)

  Popularity bias is a prominent phenomenon in recommender systems (RS), especially in the music domain. Although popularity bias mitigation techniques are known to enhance the fairness of RS while maintaining their high performance, there is a lack of understanding regarding users’ actual perception of the suggested music. To address this gap, we conducted a user study (n=40) exploring user satisfaction and perception of personalized music recommendations generated by algorithms that explicitly mitigate popularity bias. Specifically, we investigate item-centered and user-centered bias mitigation techniques, aiming to ensure fairness for artists or users, respectively. Results show that neither mitigation technique harms the users’ satisfaction with the recommendation lists despite promoting underrepresented items. However, the item-centered mitigation technique impacts user perception; by promoting less popular items, it reduces users’ familiarity with the items. Lower familiarity evokes discovery—the feeling that the recommendations enrich the user’s taste. We demonstrate that this can ultimately lead to higher satisfaction, highlighting the potential of less-popular recommendations to improve the user experience.

  Full text in ACM Digital Library

• RES 🕓15Not All Videos Become Outdated: Short-Video Recommendation by Learning to Deconfound Release Interval Bias
  by Lulu Dong (East China Normal University), Guoxiu He (East China Normal University) and Aixin Sun (Nanyang Technological University)

  Short-video recommender systems often exhibit a biased preference to recently released videos. However, not all videos become outdated; certain classic videos can still attract user’s attention. Such bias along temporal dimension can be further aggravated by the matching model between users and videos, because the model learns from preexisting interactions. From real data, we observe that different videos have varying sensitivities to recency in attracting users’ attention. Our analysis, based on a causal graph modeling short-video recommendation, suggests that the release interval serves as a confounder, establishing a backdoor path between users and videos. To address this confounding effect, we propose a model-agnostic causal architecture called Learning to Deconfound the Release Interval Bias (LDRI). LDRI enables jointly learning of the matching model and the video recency sensitivity perceptron. In the inference stage, we apply a backdoor adjustment, effectively blocking the backdoor path by intervening on each video. Extensive experiments on two benchmarks demonstrate that LDRI consistently outperforms backbone models and exhibits superior performance against state-of-the-art models. Additional comprehensive analyses confirm the deconfounding capability of LDRI.

  Full text in ACM Digital Library

Back to program