Doctoral Symposium – Algorithms & Explanations II

Date: Monday October 14
Time: 14:30-16:00
Location: Room M

• DSExplainable Multi-Stakeholder Job Recommender Systems
  by Roan Schellingerhout (Maastricht University)

  Public opinion on recommender systems has become increasingly wary in recent years. In line with this trend, lawmakers have also started to become more critical of such systems, resulting in the introduction of new laws focusing on aspects such as privacy, fairness, and explainability for recommender systems and AI at large. These concepts are especially crucial in high-risk domains such as recruitment. In recruitment specifically, decisions carry substantial weight, as the outcomes can significantly impact individuals’ careers and companies’ success. Additionally, there is a need for a multi-stakeholder approach, as these systems are used by job seekers, recruiters, and companies simultaneously, each with its own requirements and expectations. In this paper, I summarize my current research on the topic of explainable, multi-stakeholder job recommender systems and set out a number of future research directions.

• DSExplainable and Faithful Educational Recommendations through Causal Language Modelling via Knowledge Graphs
  by Neda Afreen (University of Cagliari)

  The rapid expansion of digital education has significantly increased the need for recommender systems to help learners navigate the extensive variety of available learning resources. Recent advancements in these systems have notably improved the personalization of course recommendations. However, many existing systems fail to provide clear explanations for their recommendations, making it difficult for learners to understand why a particular suggestion was made. Researchers have emphasized the importance of explanations in various domains such as e-commerce, media, and entertainment, demonstrating how explanations can enhance system transparency, foster user trust, and improve decision-making processes. Despite these insights, such approaches have been rarely applied to the educational domain, and their effectiveness in practical use remains largely unexamined. My research focuses on developing explainable recommender systems for digital education. First, I aim to design knowledge graphs that can support high-quality recommendations in the educational context. Second, I will create models backed by these knowledge graphs that not only deliver accurate recommendations but also provide faithful explanations for each suggestion. Third, I will evaluate the effectiveness of these explainable recommender systems in real-world educational environments. Ultimately, this research aims to advance the development of more transparent and user-centric educational technologies.

• DSFairness and Transparency in Music Recommender Systems: Improvements for Artists
  by Karlijn Dinnissen (Utrecht University)

  Music streaming services have become one of the main sources of music consumption in the last decade, with recommender systems playing a crucial role. Since these systems partially determine which songs listeners hear, they significantly influence the artists behind the music. However, when assessing the performance and fairness of music recommender systems, the perspectives of artists and others working in the music industry are often overlooked. Additionally, artists express a desire for greater transparency regarding why certain songs are recommended while others are not. This research project adopts a multi-stakeholder approach to close the gap between music recommender systems and the artists whose music they recommend. First, we gather insights from artists and music industry professionals through interviews and questionnaires. Building on those insights, we then aim to improve matching between end users and music from lesser-known artists by generating rich item and user representations. Results will be evaluated both quantitatively and qualitatively. Lastly, we plan to effectively communicate music recommender system fairness by increasing transparency for both end users and artists.

• DSExplainability in Music Recommender System
  by Shahrzad Shashaani (TU Wien)

  Recommendation systems play a crucial role in our daily lives, influencing many of our significant and minor decisions. These systems also have become integral to the music industry, guiding users to discover new content based on their tastes. However, the lack of transparency in these systems often leaves users questioning the rationale behind recommendations. To address this issue, adding transparency and explainability to recommender systems is a promising solution. Enhancing the explainability of these systems can significantly improve user trust and satisfaction. This research focuses on exploring transparency and explainability in the context of recommendation systems, focusing on the music domain. This research can help to understand the gaps in explainability in music recommender systems to create more engaging and trustworthy music recommendations.

Back to program