נושא הפרוייקט
מספר פרוייקט
מחלקה
שמות סטודנטים
אימייל
שמות מנחים
בעיית הפלייליסט השיתופי: אלגוריתם ופיתוח.
Multiple Attribute List Aggregation: Application to Democratic Playlist Editing
תקציר בעיברית
תקציר באנגלית
We find the subject of complicated group decision-making intriguing, as it encompasses various aspects that warrant further research. Particularly interesting are scenarios where decisions must address global constraints and involve combinatorial optimization. Our focus is currently on this topic for my thesis, on multi-attributed social choice from both voting and fair division perspective. We present a social choice model that incorporates time-based constraints, where the goal is to produce an ordered list that satisfies both agent preferences (based on approval ballots) and global constraints. First, we analyze the general model, showing that it is generally NP-hard, but admits polynomial-time algorithms for a special case; we also develop heuristic solutions for the general case. Next, we explore the potential applications of the model and demonstrate its relevance by focusing on the use case of the democratic playlist editing. In this scenario, our aim is to generate a playlist that reflects agent preferences for a given set of musical tracks while also considering soft constraints regarding the sequencing and transitions of tracks over time. We illustrate how the problem of democratic playlist editing can be translated into our model, and present simulation results where we apply our heuristics to solve specific instances of the problem. All instances were generated utilizing Spotify's API and conducting simulations based on a predefined set of rules that define the characteristics of a "good" playlist, as outlined in the existing literature. The simulations indicates that the simulated annealing heuristic performed best in finding a solution that satisfies both simulated user preferences and global constraints, while genetic algorithm and greedy algorithm performed less effectively. We contend that our results are promising, not only for the specific use case of democratic playlist editing, but also for other cooperative-use cases such as: democratic scheduling, democratic production, and participatory budgeting. An article has been submitted to EUMAS-2023 in Naples, describing this work.