Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unveiling and Simulating Short-Video Addiction Behaviors via Economic Addiction Theory

Chen Xu, Zhipeng Yi, Ruizi Wang, Wenjie Wang, Jun Xu, Maarten de Rijke

TL;DR

The paper tackles short-video addiction by framing it through economic addiction theory and leveraging large-scale implicit data. It introduces AddictSim, an LLM-based simulator trained with a mean-to-adapted (M2A) strategy and group relative policy optimization (GRPO) to capture both population-level and personalized addiction dynamics. Empirical results on THU and KuaiRec show addiction patterns resemble traditional reinforcement and tolerance effects, with significant heterogeneity across user groups; AddictSim outperforms baselines in watch-time prediction and demonstrates that diversity-aware re-ranking can mitigate addictive engagement. The work provides a scalable framework for evaluating and designing healthier recommendation systems and offers actionable insights for reducing addiction via content diversification.

Abstract

Short-video applications have attracted substantial user traffic. However, these platforms also foster problematic usage patterns, commonly referred to as short-video addiction, which pose risks to both user health and the sustainable development of platforms. Prior studies on this issue have primarily relied on questionnaires or volunteer-based data collection, which are often limited by small sample sizes and population biases. In contrast, short-video platforms have large-scale behavioral data, offering a valuable foundation for analyzing addictive behaviors. To examine addiction-aware behavior patterns, we combine economic addiction theory with users' implicit behavior captured by recommendation systems. Our analysis shows that short-video addiction follows functional patterns similar to traditional forms of addictive behavior (e.g., substance abuse) and that its intensity is consistent with findings from previous social science studies. To develop a simulator that can learn and model these patterns, we introduce a novel training framework, AddictSim. To consider the personalized addiction patterns, AddictSim uses a mean-to-adapted strategy with group relative policy optimization training. Experiments on two large-scale datasets show that AddictSim consistently outperforms existing training strategies. Our simulation results show that integrating diversity-aware algorithms can mitigate addictive behaviors well.

Unveiling and Simulating Short-Video Addiction Behaviors via Economic Addiction Theory

TL;DR

The paper tackles short-video addiction by framing it through economic addiction theory and leveraging large-scale implicit data. It introduces AddictSim, an LLM-based simulator trained with a mean-to-adapted (M2A) strategy and group relative policy optimization (GRPO) to capture both population-level and personalized addiction dynamics. Empirical results on THU and KuaiRec show addiction patterns resemble traditional reinforcement and tolerance effects, with significant heterogeneity across user groups; AddictSim outperforms baselines in watch-time prediction and demonstrates that diversity-aware re-ranking can mitigate addictive engagement. The work provides a scalable framework for evaluating and designing healthier recommendation systems and offers actionable insights for reducing addiction via content diversification.

Abstract

Short-video applications have attracted substantial user traffic. However, these platforms also foster problematic usage patterns, commonly referred to as short-video addiction, which pose risks to both user health and the sustainable development of platforms. Prior studies on this issue have primarily relied on questionnaires or volunteer-based data collection, which are often limited by small sample sizes and population biases. In contrast, short-video platforms have large-scale behavioral data, offering a valuable foundation for analyzing addictive behaviors. To examine addiction-aware behavior patterns, we combine economic addiction theory with users' implicit behavior captured by recommendation systems. Our analysis shows that short-video addiction follows functional patterns similar to traditional forms of addictive behavior (e.g., substance abuse) and that its intensity is consistent with findings from previous social science studies. To develop a simulator that can learn and model these patterns, we introduce a novel training framework, AddictSim. To consider the personalized addiction patterns, AddictSim uses a mean-to-adapted strategy with group relative policy optimization training. Experiments on two large-scale datasets show that AddictSim consistently outperforms existing training strategies. Our simulation results show that integrating diversity-aware algorithms can mitigate addictive behaviors well.
Paper Structure (29 sections, 17 equations, 10 figures, 6 tables, 1 algorithm)

This paper contains 29 sections, 17 equations, 10 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Addiction Behavior Discovery
  5. Short-video addiction mapping
  6. Short-video addiction model
  7. Experimental analysis
  8. Experimental settings
  9. Experimental results
  10. Simulator: AddictSim
  11. Economic addiction model as reward model
  12. Stage-1 Reward $R^1$
  13. Stage-2 Reward $R^2$
  14. M2A training strategy
  15. Experimental results
  16. ...and 14 more sections

Figures (10)

  • Figure 1: Addiction patterns on a popular short-video platform. The x-axis represents cumulative user watch time, and the y-axis represents user utility.
  • Figure 2: The framework of our study consists of two main steps. Step 1: construct an addiction model. Step 2: train a simulator, AddictSim, using the addiction model as the reward function, and apply AddictSim to run simulations aimed at mitigating addictive behavior.
  • Figure 3: Average addiction model parameters $a$, $b$, and $w$ for users on the THU and KuaiRec datasets. The ratio $a/b$ represents the duration of addiction, while $w$ indicates the intensity of addiction.
  • Figure 4: Comparison of addictive behaviors across different recommendation scenarios, including short-video, hybrid, and traditional e-commerce settings.
  • Figure 5: Addictive behaviors among different user groups.
  • ...and 5 more figures