Proactive Guiding Strategy for Item-side Fairness in Interactive Recommendation

Chongjun Xia; Xiaoyu Shi; Hong Xie; Xianzhi Wang; yun lu; Mingsheng Shang

cs.IRMar 3, 2026

Proactive Guiding Strategy for Item-side Fairness in Interactive Recommendation

Authors: Chongjun Xia, Xiaoyu Shi, Hong Xie, Xianzhi Wang, yun lu, Mingsheng Shang

Organizations: Chinese Academy of Sciences · University of Technology Sydney · University of Science and Technology of China

Abstract

Item-side fairness is crucial for ensuring the fair exposure of long-tail items in interactive recommender systems. Existing approaches promote the exposure of long-tail items by directly incorporating them into recommended results. This causes misalignment between user preferences and the recommended long-tail items, which hinders long-term user engagement and reduces the effectiveness of recommendations. We aim for a proactive fairness-guiding strategy, which actively guides user preferences toward long-tail items while preserving user satisfaction during the interactive recommendation process. To this end, we propose HRL4PFG, an interactive recommendation framework that leverages hierarchical reinforcement learning to guide user preferences toward long-tail items progressively. HRL4PFG operates through a macro-level process that generates fairness-guided targets based on multi-step feedback, and a micro-level process that fine-tunes recommendations in real time according to both these targets and evolving user preferences. Extensive experiments show that HRL4PFG improves cumulative interaction rewards and maximum user interaction length by a larger margin when compared with state-of-the-art methods in interactive recommendation environments.

Figures & tables

Figure 1: Illustration of two types of item-side fairness-aware recommendation.

Figure 2: The overall architecture of HRL4PFG.

Dataset	Usage	Users	Items	Interactions
KuaiRec	Train	7,176	10,728	12530.8k
KuaiRec	Test	1,411	3,327	4676.5k
KuaiRand	Train	26,285	7,551	1436.6k
KuaiRand	Test	27,285	7,583	1186.1k

Table 1: Dataset statistics

Envs	Methods	Max Len 30				Max Len 50
Envs	Methods	$\text{R}_{\text{cum}}$ $\uparrow$	$\text{R}_{\text{single}}$ $\uparrow$	Len $\uparrow$	Gini Index (%) $\downarrow$	$\text{R}_{\text{cum}}$ $\uparrow$	$\text{R}_{\text{single}}$ $\uparrow$	Len $\uparrow$	Gini Index (%) $\downarrow$
KuaiRec	SQN	13.46 $\pm$ 1.96	1.00 $\pm$ 0.03	13.61 $\pm$ 2.35	99.40 $\pm$ 0.10	13.23 $\pm$ 4.32	0.99 $\pm$ 0.04	13.65 $\pm$ 5.13	99.30 $\pm$ 0.10
	PG	14.06 $\pm$ 1.23	0.79 $\pm$ 0.08	18.32 $\pm$ 2.95	99.10 $\pm$ 0.20	17.76 $\pm$ 1.60	0.71 $\pm$ 0.10	26.02 $\pm$ 5.19	98.70 $\pm$ 0.20
	DDPG	8.23 $\pm$ 3.99	0.80 $\pm$ 0.22	13.03 $\pm$ 8.59	99.50 $\pm$ 0.20	12.27 $\pm$ 4.41	0.42 $\pm$ 0.17	32.92 $\pm$ 10.29	99.40 $\pm$ 0.30
	TD3	8.27 $\pm$ 5.68	0.44 $\pm$ 0.28	23.13 $\pm$ 8.29	99.60 $\pm$ 0.30	9.34 $\pm$ 6.69	0.50 $\pm$ 0.34	31.39 $\pm$ 19.43	99.30 $\pm$ 0.60
	C51	13.90 $\pm$ 3.38	0.77 $\pm$ 0.09	18.84 $\pm$ 6.03	99.00 $\pm$ 0.20	19.12 $\pm$ 5.50	0.76 $\pm$ 0.06	25.61 $\pm$ 8.47	98.40 $\pm$ 0.30

Table 2: Average results of all methods in different environments.

Figure 3: Results under different exit conditions.

Figure 4: Ablation study results of HRL4PFG.

Figure 5: Hyper-parameter study on two datasets.

Long-Tail Recommendation

See all 8 papers

Fairness Begins with State: Purifying Latent Preferences for Hierarchical Reinforcement Learning in Interactive Recommendation
Mar 4, 2026Yun Lu, Xiaoyu Shi, Hong Xie +2Fairness in Recommender Systems Recommender Systems
Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences Chongqing School, University of Chinese Academy of Sciences Chongqing, China · The First Affiliated Hospital, University of Science and Technology of China Hefei, China · City University of Hong Kong Hong Kong, Hong Kong
Taming the Long Tail: Efficient Item-wise Sharpness-Aware Minimization for LLM-based Recommender Systems
Mar 13, 2026Jiaming Zhang, Yuyuan Li, Xiaohua Feng +4Long-Tail Recommendation Recommender Systems
Zhejiang University Hangzhou, Zhejiang, China · Hangzhou Dianzi University Hangzhou, Zhejiang, China · Ant Group Hangzhou, Zhejiang, China
Modeling User Exploration Saturation: When Recommender Systems Should Stop Pushing Novelty
Apr 2, 2026Enock O. Ayiku, Evelyn Osei, Emebo OnyekaRecommender Systems Fairness in Recommender Systems
University of Massachusetts Boston, Boston, MA, USA · Kansas State University, Manhattan, KS, USA · Virginia Tech, Blacksburg, VA, USA

Mar 4, 2026J/K move · Enter open · S save

Fairness Begins with State: Purifying Latent Preferences for Hierarchical Reinforcement Learning in Interactive Recommendation

Yun Lu, Xiaoyu Shi, Hong Xie, Xiangyu Zhao+1

Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences Chongqing School, University of Chinese Academy of Sciences Chongqing, China · The First Affiliated Hospital, University of Science and Technology of China Hefei, China · City University of Hong Kong Hong Kong, Hong Kong

Fairness in Recommender Systems Recommender Systems Algorithmic Fairness User Preference Modeling

Open paper AskPDF arXiv

Interactive recommender systems (IRS) are increasingly optimized with Reinforcement Learning (RL) to capture the sequential nature of user-system dynamics. However, existing fairness-aware methods often suffer from a fundamental oversight: they assume the observed user state is a faithful representation of true preferences. In reality, implicit feedback is contaminated by popularity-driven noise and exposure bias, creating a distorted state that misleads the RL agent. We argue that the persistent conflict between accuracy and fairness is not merely a reward-shaping issue, but a state estimation failure. In this work, we propose DSRM-HRL, a framework that reformulates fairness-aware recommendation as a latent state purification problem followed by decoupled hierarchical decision-making. We introduce a Denoising State Representation Module (DSRM) based on diffusion models to recover the low-entropy latent preference manifold from high-entropy, noisy interaction histories. Built upon this purified state, a Hierarchical Reinforcement Learning (HRL) agent is employed to decouple conflicting objectives: a high-level policy regulates long-term fairness trajectories, while a low-level policy optimizes short-term engagement under these dynamic constraints. Extensive experiments on high-fidelity simulators (KuaiRec, KuaiRand) demonstrate that DSRM-HRL effectively breaks the "rich-get-richer" feedback loop, achieving a superior Pareto frontier between recommendation utility and exposure equity.

Abstract

Long-Tail Recommendation

Explore similar work

Fairness Begins with State: Purifying Latent Preferences for Hierarchical Reinforcement Learning in Interactive Recommendation

Taming the Long Tail: Efficient Item-wise Sharpness-Aware Minimization for LLM-based Recommender Systems

Modeling User Exploration Saturation: When Recommender Systems Should Stop Pushing Novelty