Doctoral Symposium – Societal RecSys I

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

• DSHow to Evaluate Serendipity in Recommender Systems: the Need for a Serendiptionnaire
  by Brett Binst (imec-SMIT, Vrije Universiteit Brussel)

  Recommender systems can assist in various user tasks and serve diverse values, including exploring the item space. Serendipity has recently received considerable attention, often seen as a way to broaden users’ tastes and counteract filter bubbles. However, the field of research on serendipity is fragmented regarding its evaluation methods, which impedes the progress of knowledge accumulation. This research plan proposes two studies to address these issues. First, a systematic literature review will be conducted to provide insights into how serendipity is currently studied in the field. This review will serve as a reference for novice researchers and help mitigate fragmentation by presenting a thorough overview of the field. This systematic literature review has already revealed a significant gap: the lack of a validated, widely accepted method for evaluating serendipity. Therefore, the second part of this research plan is to develop a validated questionnaire, the serendiptionnaire, to measure serendipity. This tool will provide a ground truth for evaluating serendipity, aiding in answering fundamental questions within the field and validating offline metrics.

• DSPersonal Values and Community-Centric Environmental Recommender Systems: Enhancing Sustainability Through User Engagement
  by Bianca Maria Deconcini (University of Turin)

  The concept of sustainability has become a central focus across multiple sectors, driven by the urgent need to address climate change and protect the environment. Technological advancements and capabilities, together with the emergence of new ecological issues [25], are leading to growing awareness and influencing shifts in multiple areas such as energy, transportation, and waste management. Within this context, the roles of recommender systems represent a promising solution, since people need guidance and occasionally a gentle push to translate their intentions into actions or to bring goals to life [9]. However, existing literature reveals a fragmented landscape, with solutions often addressing specific aspects or recommendation contribution in isolation. Many sustainability interventions focus solely on providing consumption data and environmental insights, while others emphasize learning and behavior change strategies. My doctoral project aims to address this gap by leveraging various approaches to recommender systems and applying them in sustainability contexts, with the goal to build a holistic system that maximizes the contributions of these diverse methods, also integrating user-centric and value-driven perspectives. This research project delves into two distinct facets: energy sustainability and sustainable mobility. The first case centers on enhancing energy efficiency within energy communities through personalized recommendations and engagement strategies. The second facet focuses on reshaping user commuting patterns towards sustainable alternatives, by recommending suitable and more sustainable modes of transportation, such as cycling, carpooling, and public transportation. Both cases share the same objective: align user behaviors with sustainability goals, thereby reducing individual environmental impact and enhancing the sense of belonging to a community, whether this is confined to a group of individuals or pertains to society at large. An innovative comprehensive recommendation system approach is highly beneficial since it can take advantage of all the existing contributions combined in a framework that makes at the same time different types of recommendations: explainable, educative, behavioral and social-aware, addressing the complexities of this multifaceted domain.

• DSEnhancing Privacy in Recommender Systems through Differential Privacy Techniques
  by Angela Di Fazio (Politecnico di Bari)

  Recommender systems have become essential tools for addressing information overload in the digital age. However, the collection and usage of user data for personalized recommendations raise significant privacy concerns. This research focuses on enhancing privacy in recommender systems through the application of differential privacy techniques, particularly in the domain of privacy-preserving data publishing.

  Our study aims to address three key research questions: (1) developing standardized metrics to characterize and compare recommendation datasets in the context of privacy-preserving data publishing, (2) designing differential privacy algorithms for private data publishing that preserve recommendation quality, and (3) examining the impact of differential privacy on beyond-accuracy objectives in recommender systems.

  We propose to develop domain-specific metrics for evaluating the similarity between recommendation datasets, analogous to those used in other domains such as trajectory data publication. Additionally, we will investigate methods to balance the trade-off between privacy guarantees and recommendation accuracy, considering the potential disparate impacts on different user subgroups. Finally, we aim to assess the broader implications of implementing differential privacy on beyond-accuracy objectives such as diversity, popularity bias, and fairness.

  By addressing these challenges, our research seeks to contribute to the advancement of privacy-preserving techniques in recommender systems, facilitating the responsible and secure use of recommendation data while maintaining the utility of personalized suggestions. The outcomes of this study have the potential to significantly benefit the field by enabling the reuse of existing algorithms with minimal adjustments while ensuring robust privacy guarantees.

• DSA New Perspective in Health Recommendations: Integration of Human Pose Estimation
  by Gaetano Dibenedetto (University of Bari Aldo Moro)

  In recent years, there has been a growing interest in multimodal and multi-source data due to their ability to introduce heterogeneous information. Studies have demonstrated that combining such information enhances the performance of Recommender Systems across various scenarios. In the context of Health Recommendation Systems (HRS), different types of data are utilized, primarily focusing on patient-based information, but data from Pose Estimations (PE) are not incorporated.

  The objective of my Ph.D. is to investigate methods to design and develop HRS that treat the PE as one of the input sources, taking into account aspects such as privacy concerns and balancing the trade-off between system quality and responsiveness. By leveraging the combination of diverse information sources, I intend to create a new model in the area of HRS capable of providing more precise and explainable recommendations.

Back to program