Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

User Welfare Optimization in Recommender Systems with Competing Content Creators

Fan Yao, Yiming Liao, Mingzhe Wu, Chuanhao Li, Yan Zhu, James Yang, Qifan Wang, Haifeng Xu, Hongning Wang

TL;DR

This work models user welfare optimization in recommender systems with competing content creators as an extended Content Creator Competition game, $C^3_{ext}$, and addresses suboptimal outcomes caused by creators' local, information-limited updates. It proposes three platform intervention mechanisms—User Importance Reweighting (UIR), Soft Matching Truncation (SMT), and Hard Matching Truncation (HMT)—and an adaptive reweighting algorithm that uses multiplicative weight updates to steer attention toward underserved users, thereby improving global welfare $W(s)$. The approach is validated through offline simulations on synthetic and MovieLens-1m data, showing welfare gains and varying effects on group utilities and diversity, with HMT delivering strong, stable improvements. An online three-week deployment on a leading platform confirms increased engagement and content diversity, illustrating practical feasibility and potential impact for incentive-aware recommender systems.

Abstract

Driven by the new economic opportunities created by the creator economy, an increasing number of content creators rely on and compete for revenue generated from online content recommendation platforms. This burgeoning competition reshapes the dynamics of content distribution and profoundly impacts long-term user welfare on the platform. However, the absence of a comprehensive picture of global user preference distribution often traps the competition, especially the creators, in states that yield sub-optimal user welfare. To encourage creators to best serve a broad user population with relevant content, it becomes the platform's responsibility to leverage its information advantage regarding user preference distribution to accurately signal creators. In this study, we perform system-side user welfare optimization under a competitive game setting among content creators. We propose an algorithmic solution for the platform, which dynamically computes a sequence of weights for each user based on their satisfaction of the recommended content. These weights are then utilized to design mechanisms that adjust the recommendation policy or the post-recommendation rewards, thereby influencing creators' content production strategies. To validate the effectiveness of our proposed method, we report our findings from a series of experiments, including: 1. a proof-of-concept negative example illustrating how creators' strategies converge towards sub-optimal states without platform intervention; 2. offline experiments employing our proposed intervention mechanisms on diverse datasets; and 3. results from a three-week online experiment conducted on a leading short-video recommendation platform.

User Welfare Optimization in Recommender Systems with Competing Content Creators

TL;DR

This work models user welfare optimization in recommender systems with competing content creators as an extended Content Creator Competition game, , and addresses suboptimal outcomes caused by creators' local, information-limited updates. It proposes three platform intervention mechanisms—User Importance Reweighting (UIR), Soft Matching Truncation (SMT), and Hard Matching Truncation (HMT)—and an adaptive reweighting algorithm that uses multiplicative weight updates to steer attention toward underserved users, thereby improving global welfare . The approach is validated through offline simulations on synthetic and MovieLens-1m data, showing welfare gains and varying effects on group utilities and diversity, with HMT delivering strong, stable improvements. An online three-week deployment on a leading platform confirms increased engagement and content diversity, illustrating practical feasibility and potential impact for incentive-aware recommender systems.

Abstract

Driven by the new economic opportunities created by the creator economy, an increasing number of content creators rely on and compete for revenue generated from online content recommendation platforms. This burgeoning competition reshapes the dynamics of content distribution and profoundly impacts long-term user welfare on the platform. However, the absence of a comprehensive picture of global user preference distribution often traps the competition, especially the creators, in states that yield sub-optimal user welfare. To encourage creators to best serve a broad user population with relevant content, it becomes the platform's responsibility to leverage its information advantage regarding user preference distribution to accurately signal creators. In this study, we perform system-side user welfare optimization under a competitive game setting among content creators. We propose an algorithmic solution for the platform, which dynamically computes a sequence of weights for each user based on their satisfaction of the recommended content. These weights are then utilized to design mechanisms that adjust the recommendation policy or the post-recommendation rewards, thereby influencing creators' content production strategies. To validate the effectiveness of our proposed method, we report our findings from a series of experiments, including: 1. a proof-of-concept negative example illustrating how creators' strategies converge towards sub-optimal states without platform intervention; 2. offline experiments employing our proposed intervention mechanisms on diverse datasets; and 3. results from a three-week online experiment conducted on a leading short-video recommendation platform.
Paper Structure (21 sections, 3 theorems, 30 equations, 3 figures, 2 tables, 5 algorithms)

This paper contains 21 sections, 3 theorems, 30 equations, 3 figures, 2 tables, 5 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. The Modeling of Content Creation Competition
  4. Intervention Mechanism Design
  5. The Necessity of Intervention
  6. Platform's Intervention Mechanisms
  7. Welfare Optimization through Adaptive Reweighing
  8. Experiments
  9. Experiments on Offline Data
  10. Synthetic environment
  11. Environment constructed from MovieLens-1m
  12. Configurations of adaptive reweighting algorithm and intervention mechanisms
  13. Results
  14. Online Experiments
  15. Experiment Setups
  16. ...and 6 more sections

Key Result

Theorem 1

Any $C^{3}_{\text{ext}}$ game with $K=n$ has a unique pure Nash equilibrium (PNE) under the utility function eq:creator_expected_utility_func2 if $\sigma(\cdot)$ is sufficiently smooth and concave and each creator has a convex strategy set.

Figures (3)

  • Figure 1: Visualization of creators' evolving strategies. Left: no intervention, right: platform decreases the weight of the center user by half. Creators' strategies are marked with different colors, and the arrows start from initial strategies and point to the last-iterate strategies.
  • Figure 2: Performance of UIR, SMT and HMT on synthetic dataset against the no-intervention baseline. Results are averaged over 10 independently sampled synthetic environments including one-sigma error bars. $x$-axis: group sizes divided by 10.
  • Figure 3: Performance of UIR, SMT and HMT on MovieLens-1m dataset against the no-intervention baseline. Results are averaged over 10 independent simulations including 0.2-sigma error bars.

Theorems & Definitions (4)

  • Theorem 1
  • Definition 1
  • Theorem 2
  • Theorem 3