Contextual Scalarisation Thompson Sampling for multi-objective decisions in public media

2026-05-29

AI Summary

Public media recommendation must dynamically balance multiple objectives—such as audience reach, cultural value, and public service—but existing approaches rely on fixed weights or static Pareto optimization, limiting their adaptability to changing contexts. This work proposes Contextual Scalarized Thompson Sampling (CSTS), which introduces a learnable, context-aware scalarization mechanism into a multi-objective contextual bandit framework, enabling objective weights to adaptively adjust according to situational context in alignment with expert curation logic. Experiments on real program data from Swiss national broadcasting demonstrate that CSTS significantly outperforms both fixed-weight and standard contextual bandit methods, enhancing both contextual relevance of recommendations and consistency with expert-curated selections.

Abstract

Recommender systems may operate under multiple, competing objectives. For example, audience reach, cultural values, public service mandate, and operational constraints must be balanced in editorial decisions of public service media. Existing approaches relying on fixed combinations of objectives or Pareto-based optimisation do not adapt to changing priorities across situations. In this paper, we propose Contextual Scalarisation Thompson Sampler (CSTS), a multi-objective contextual bandit method that learns to weight objectives as a function of the observed context. We evaluate CSTS on real programming data from Radio Télévision Suisse, the Swiss national broadcaster, showing improved contextual relevance and better alignment with expert curation practices compared to fixed weight and standard contextual bandit approaches.

Problem

• multi-objective decision
• contextual bandits
• public media
• objective weighting
• recommendation systems

Innovation

• Contextual Scalarisation
• Thompson Sampling
• Multi-objective Bandits
• Adaptive Weighting
• Public Media Recommender Systems