Doctoral Symposium – Algorithms & Explanations I

Date: Monday October 14
Time: 11:15-12:45
Location: Room M

• DSSupporting Knowledge Workers through Personal Information Assistance with Context-aware Recommender Systems
  by Mahta Bakhshizadeh (German Research Center for Artificial Intelligence – DFKI)

  Recommender systems are extensively employed across various domains to mitigate information overload by providing personalized content. Despite their widespread use in sectors such as streaming, e-commerce, and social networks, utilizing them for personal information assistance is a comparatively novel application. This emerging application aims to develop intelligent systems capable of proactively providing knowledge workers with the most relevant information based on their context to enhance productivity. In this paper, we explore this innovative application by first defining the scope of our study, outlining the key objectives, and introducing the main challenges. We then present our current results and progress, including a comprehensive literature review, the proposal of a framework, the collection of a pioneering dataset, and the establishment of a benchmark for evaluating a recommendation scenario on our published dataset. We also discuss our ongoing efforts and future research directions.

• DSIntegrating Matrix Factorization with Graph based Models
  by Rachana Mehta (Pandit Deendayal Energy University)

  Graph based Recommender models make use of user-item rating and user-user social relationships to elicit recommendation performance by extracting inherent geometrical knowledge. In a social graph scenario, user-user trust plays a significant role in reducing sparsity and has varied characteristics that can be exploited. Existing models limit themselves to learning from either a high-order interaction graph of user-item ratings or a user-user social graph from trust value. They explore other trust characteristics in a very limited setting. The graph based model, designed using entire user-user social information, impacts performance and escalates complexities in model learning. To alleviate these issues of graph learning, graph recommender seeks assistance from matrix factorization techniques. Incorporating graph based model with matrix factorization brings its own set of challenges of model integration, leveraging trust, graph learning, and optimization. This article presents the existing work in that line and future possibilities and challenges to be catered to through novel developments

• DSFairness Explanations in Recommender Systems
  by Luan Souza (University of São Paulo)

  Fairness in recommender systems is an emerging area that aims to study and mitigate discriminations against individuals or/and groups of individuals in recommendation engines. These mitigation strategies rely on bias detection, which is a non-trivial task that requires complex analysis and interventions to ensure fairness in these engines. Furthermore, fairness interventions in recommender systems involve a trade-off between fairness and performance of the recommendation lists, impacting the user experience with less potentially accurate lists. In this context, fairness interventions with explanations have been proposed recently in the literature, mitigating discrimination in recommendation lists and providing explainability about the recommendation process and the impact of the fairness interventions in the outcomes. However, in spite of the different approaches it is still not clear how these proposals compare with each other, even those that propose to mitigate the same kind of bias. In addition, the contribution of these different explainable algorithmic fairness approaches to users’ fairness perceptions was not explored until the moment. Looking at these gaps, our doctorate project aims to investigate how these explainable fairness proposals compare to each other and how they are perceived by the users, in order to identify which fairness interventions and explanation strategies are most promising to increase transparency and fairness perceptions of recommendation lists.

Back to program