The Bandit's Blind Spot: The Critical Role of User State Representation in Recommender Systems

TL;DR

This paper highlights the importance of user state representation in recommender systems using contextual bandits, showing that embedding quality significantly impacts performance and varies across datasets.

Contribution

It systematically evaluates how different embedding-based user state representations affect bandit algorithm performance, revealing the critical role of state construction.

Findings

Variations in user state embeddings can outperform changes in bandit algorithms.

No single embedding strategy is best across all datasets.

Emphasizes the need for domain-specific evaluation of state representations.

Abstract

With the increasing availability of online information, recommender systems have become an important tool for many web-based systems. Due to the continuous aspect of recommendation environments, these systems increasingly rely on contextual multi-armed bandits (CMAB) to deliver personalized and real-time suggestions. A critical yet underexplored component in these systems is the representation of user state, which typically encapsulates the user's interaction history and is deeply correlated with the model's decisions and learning. In this paper, we investigate the impact of different embedding-based state representations derived from matrix factorization models on the performance of traditional CMAB algorithms. Our large-scale experiments reveal that variations in state representation can lead to improvements greater than those achieved by changing the bandit algorithm itself. Furthermore, no single embedding or aggregation strategy consistently dominates across datasets, underscoring the need for domain-specific evaluation. These results expose a substantial gap in the literature and emphasize that advancing bandit-based recommender systems requires a holistic approach that prioritizes embedding quality and state construction alongside algorithmic innovation. The source code for our experiments is publicly available on https://github.com/UFSCar-LaSID/bandits_blind_spot.