Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unveiling User Satisfaction and Creator Productivity Trade-Offs in Recommendation Platforms

Fan Yao, Yiming Liao, Jingzhou Liu, Shaoliang Nie, Qifan Wang, Haifeng Xu, Hongning Wang

TL;DR

This work demonstrates that a purely relevance-driven policy with low exploration strength boosts short-term user satisfaction but undermines the long-term richness of the content pool, and proposes an efficient optimization method to identify the optimal exploration strength, balancing user and creator engagement.

Abstract

On User-Generated Content (UGC) platforms, recommendation algorithms significantly impact creators' motivation to produce content as they compete for algorithmically allocated user traffic. This phenomenon subtly shapes the volume and diversity of the content pool, which is crucial for the platform's sustainability. In this work, we demonstrate, both theoretically and empirically, that a purely relevance-driven policy with low exploration strength boosts short-term user satisfaction but undermines the long-term richness of the content pool. In contrast, a more aggressive exploration policy may slightly compromise user satisfaction but promote higher content creation volume. Our findings reveal a fundamental trade-off between immediate user satisfaction and overall content production on UGC platforms. Building on this finding, we propose an efficient optimization method to identify the optimal exploration strength, balancing user and creator engagement. Our model can serve as a pre-deployment audit tool for recommendation algorithms on UGC platforms, helping to align their immediate objectives with sustainable, long-term goals.

Unveiling User Satisfaction and Creator Productivity Trade-Offs in Recommendation Platforms

TL;DR

This work demonstrates that a purely relevance-driven policy with low exploration strength boosts short-term user satisfaction but undermines the long-term richness of the content pool, and proposes an efficient optimization method to identify the optimal exploration strength, balancing user and creator engagement.

Abstract

On User-Generated Content (UGC) platforms, recommendation algorithms significantly impact creators' motivation to produce content as they compete for algorithmically allocated user traffic. This phenomenon subtly shapes the volume and diversity of the content pool, which is crucial for the platform's sustainability. In this work, we demonstrate, both theoretically and empirically, that a purely relevance-driven policy with low exploration strength boosts short-term user satisfaction but undermines the long-term richness of the content pool. In contrast, a more aggressive exploration policy may slightly compromise user satisfaction but promote higher content creation volume. Our findings reveal a fundamental trade-off between immediate user satisfaction and overall content production on UGC platforms. Building on this finding, we propose an efficient optimization method to identify the optimal exploration strength, balancing user and creator engagement. Our model can serve as a pre-deployment audit tool for recommendation algorithms on UGC platforms, helping to align their immediate objectives with sustainable, long-term goals.

Paper Structure

This paper contains 18 sections, 4 theorems, 45 equations, 4 figures, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. The Cournot Content Creation Competition
  4. Research questions:
  5. The PNEs of $C^4$ Games and Their Properties
  6. The Trade-Off Between User and Creator Engagement
  7. Finding the Optimal Trade-off through Optimization
  8. The Derivation of Exact Gradient
  9. Optimization with Approximated Gradients
  10. Experiments
  11. The Empirical Trade-Offs Between $U$ and $V$
  12. The Optimal PPM$(\beta)$ Found by Algorithm \ref{['algo:approx_g']}
  13. Conclusion
  14. Acknowledgment.
  15. Omitted Proofs
  16. ...and 3 more sections

Key Result

Theorem 1

For any $C^4$ instance $\mathcal{G}$$(n, m, M, \{c_i\}_{i=1}^n, \bm{\beta})$. If each $c_i$ is convex in $x_i$, $\mathcal{G}$ admits a unique PNE.

Figures (4)

  • Figure 1: The left and the middle panel: the empirical distributions of content creation frequency $x_i^*$ and each user's individual utility $\pi_j^*$. Different colors represent results for PNEs induced by different $\beta$. Right: the total content creation $V$ and total user satisfaction $U$ obtained under different $\beta$. Error bars obtained from 10 independently generated environments.
  • Figure 2: Panel 1,2: social welfare improving curve under Algorithm \ref{['algo:approx_g']}, and the distribution of the obtained optimal $\beta_j$ in the synthetic environment. Panel 3,4: the same plots in the MovieLens environment. $\lambda=0.5$.
  • Figure 3: The left and the middle panel: the empirical distributions of content creation frequency $x_i^*$ and each user's individual utility $\pi_j^*$. Different colors represent results for PNEs induced by different $\beta$. Right: the total content creation $V$ and total user satisfaction $U$ obtained under different $\beta$. Error bars obtained from 10 independently generated environments.
  • Figure 4: Panel 1,2: social welfare improving curve under Algorithm \ref{['algo:approx_g']}, and the distribution of the obtained optimal $\beta_j$ in the synthetic environment. Panel 3,4: the same plots in the MovieLens environment. $\lambda=0.1$.

Theorems & Definitions (6)

  • Definition 1
  • Theorem 1
  • Corollary 1
  • Definition 2
  • Theorem 2
  • Proposition 1