SRSUPM: Sequential Recommender System Based on User Psychological Motivation

TL;DR

This work addresses the challenge of modeling evolving user motivations in sequential recommendation by introducing SRSUPM, a plug-in framework that explicitly handles psychological motivation shift. It defines PMSA to quantify shift degrees, Shift Information Construction to create multi-level shift representations, PMSID to decompose and regularize these representations, and PMSIM to enforce shift-aware collaborative learning via InfoNCE. By integrating with diverse backbones (e.g., SASRec, GRU4Rec, Caser, Bert4Rec, LightSANs), SRSUPM yields consistent improvements on three real-world datasets and provides insights into shift distributions and robustness. The approach offers a practical path to distribution-aware, shift-sensitive recommendations with moderate inference overhead and strong empirical support.

Abstract

Sequential recommender infers users' evolving psychological motivations from historical interactions to recommend the next preferred items. Most existing methods compress recent behaviors into a single vector and optimize it toward a single observed target item, but lack explicit modeling of psychological motivation shift. As a result, they struggle to uncover the distributional patterns across different shift degrees and to capture collaborative knowledge that is sensitive to psychological motivation shift. We propose a general framework, the Sequential Recommender System Based on User Psychological Motivation, to enhance sequential recommenders with psychological motivation shift-aware user modeling. Specifically, the Psychological Motivation Shift Assessment quantitatively measures psychological motivation shift; guided by PMSA, the Shift Information Construction models dynamically evolving multi-level shift states, and the Psychological Motivation Shift-driven Information Decomposition decomposes and regularizes representations across shift levels. Moreover, the Psychological Motivation Shift Information Matching strengthens collaborative patterns related to psychological motivation shift to learn more discriminative user representations. Extensive experiments on three public benchmarks show that SRSUPM consistently outperforms representative baselines on diverse sequential recommender tasks.

Figures (8)

• Figure 1: When predicting the next interacted item, sequential recommender methods that follow the standard formulation often fail to make accurate predictions for users $m_1$ and $m_3$, mainly due to the shift in psychological motivation.
• Figure 2: Overview of SRSUPM with SASRec backbone and $V{=}5$. PMSA assesses the psychological motivation shift level $b{=}4$ between the input sequence and target item using multi-label categories. Shift Information Construction generates $V$ transformation vectors, selects the corresponding shift-level representation, and aligns it with users of the same shift level via Psychological Motivation Shift Information Matching before scoring candidate items.
• Figure 3: Performance w.r.t different values of $V$.
• Figure 4: Performance w.r.t different values of weight parameter $\gamma_1$.
• Figure 5: Performance w.r.t different values of weight parameter $\gamma_2$.