Trajectory-Based Recommender Systems as Control Systems

2026-06-22

AI Summary

Existing trajectory-based recommender systems lack a unified theoretical framework, making it difficult to effectively model user behavior paths guided by long-term goals. This work addresses this gap by systematically introducing control theory into the field of recommender systems for the first time, proposing a unified formal framework centered on the concept of “trajectory.” Within this framework, the recommendation process is modeled as a controlled dynamical system that explicitly captures the evolution of user behavior under the influence of long-term objectives. The study not only establishes a foundational theoretical formulation for trajectory-based recommendation but also demonstrates the framework’s expressiveness and applicability through empirical validation in an educational recommendation scenario, thereby opening a new research paradigm for goal-oriented, long-horizon recommendation tasks.

Abstract

Recommender Systems (RS) are a key research domain and play an increasing role in our content-overwhelmed lives. In this paper, we explore Trajectory-Based Recommender Systems (TBRS), a subfield for which many related studies exist, yet still lacking a common framework. We argue that Control Theory provides an appropriate foundation for formalizing and solving TBRS problems. TBRS, sometimes named Long Term goal Recommender Systems, share core principles with classical RS, but at their core lies the concept of a trajectory, a defining element that makes these systems a singular category. To date, most RSs that include a notion of goal or long-term objective, when this goal is explicit, have not been recognized as having specific characteristics that make them worth regrouping under a dedicated field of research. We review related work, observe how they differ from already conceptualized RSs, and sketch the foundations of a possible theoretical framework based on control theory. Finally, we show how Educational Recommender Systems (ERS), intrinsically long-term and goal-driven, can be modeled within the proposed TBRS framework.

Problem

• Trajectory-Based Recommender Systems
• Long Term Goal
• Control Theory
• Recommender Systems
• Educational Recommender Systems

Innovation

• Trajectory-Based Recommender Systems
• Control Theory
• Long-Term Goal
• Educational Recommender Systems
• Recommendation Framework