Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Artificial Intelligence and Algorithmic Price Collusion in Two-sided Markets

Cristian Chica, Yinglong Guo, Gilad Lerman

TL;DR

The paper addresses algorithmic price collusion in two-sided markets by modeling AI-driven platforms using $Q$-learning within a baseline two-stage game and an infinite-horizon repeated game framework. It demonstrates that network externalities captured by $\boldsymbol{\Phi}$ amplify tacit collusion, with higher discount rates and lower user heterogeneity further increasing collusion, while stronger outside options mitigate it. A comprehensive MARL simulation with discretized prices reveals that collusion can exceed traditional Bertrand benchmarks, and the authors propose incorporating a penalty term in the $Q$-learning update to curb collusive outcomes. The work provides policy-relevant insights for regulatory design and highlights how algorithmic pricing in markets like online gaming, streaming, and ride-sharing can leverage network effects to raise profits, underscoring the need for targeted safeguards.

Abstract

Algorithmic price collusion facilitated by artificial intelligence (AI) algorithms raises significant concerns. We examine how AI agents using Q-learning engage in tacit collusion in two-sided markets. Our experiments reveal that AI-driven platforms achieve higher collusion levels compared to Bertrand competition. Increased network externalities significantly enhance collusion, suggesting AI algorithms exploit them to maximize profits. Higher user heterogeneity or greater utility from outside options generally reduce collusion, while higher discount rates increase it. Tacit collusion remains feasible even at low discount rates. To mitigate collusive behavior and inform potential regulatory measures, we propose incorporating a penalty term in the Q-learning algorithm.

Artificial Intelligence and Algorithmic Price Collusion in Two-sided Markets

TL;DR

The paper addresses algorithmic price collusion in two-sided markets by modeling AI-driven platforms using -learning within a baseline two-stage game and an infinite-horizon repeated game framework. It demonstrates that network externalities captured by amplify tacit collusion, with higher discount rates and lower user heterogeneity further increasing collusion, while stronger outside options mitigate it. A comprehensive MARL simulation with discretized prices reveals that collusion can exceed traditional Bertrand benchmarks, and the authors propose incorporating a penalty term in the -learning update to curb collusive outcomes. The work provides policy-relevant insights for regulatory design and highlights how algorithmic pricing in markets like online gaming, streaming, and ride-sharing can leverage network effects to raise profits, underscoring the need for targeted safeguards.

Abstract

Algorithmic price collusion facilitated by artificial intelligence (AI) algorithms raises significant concerns. We examine how AI agents using Q-learning engage in tacit collusion in two-sided markets. Our experiments reveal that AI-driven platforms achieve higher collusion levels compared to Bertrand competition. Increased network externalities significantly enhance collusion, suggesting AI algorithms exploit them to maximize profits. Higher user heterogeneity or greater utility from outside options generally reduce collusion, while higher discount rates increase it. Tacit collusion remains feasible even at low discount rates. To mitigate collusive behavior and inform potential regulatory measures, we propose incorporating a penalty term in the Q-learning algorithm.
Paper Structure (20 sections, 33 equations, 19 figures, 3 tables)

This paper contains 20 sections, 33 equations, 19 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Review of Our Economic Framework
  3. The Baseline Platform Competition Game
  4. The Infinite Repeated Game
  5. Simulation Framework
  6. Preliminaries: Multi-agent Reinforcement Learning
  7. The Simulation Setup
  8. Experimental Results
  9. Dependence of the Collusive Level on the Network Externalities
  10. A Study of the Collusive Level under Special Network Externalities
  11. A Study of the Collusive Level under Special Market Parameters
  12. Discussion of Exceptional Cases
  13. Economic and Policy Discussion
  14. Economic Discussion
  15. Policy Discussion
  16. ...and 5 more sections

Figures (19)

  • Figure 1: Demonstration of the infinite repeated game framework
  • Figure 2: Demonstration of the dependence of the four fitted univariate functions on the externalities. Top left: $f_{bb}(\phi_{bb})$; Top right: $f_{ss}(\phi_{ss})$; Bottom left: $f_{bs}(\phi_{bs})$; Bottom right: $f_{sb}(\phi_{sb})$.
  • Figure 3: Heatmaps of the fitted functions $f_{bb,ss}(\phi_{bb}, \phi_{ss})$ (left) and $f_{sb,bs}(\phi_{bs}, \phi_{sb})$ (right), which capture the bivariate effect between $\phi_{bb}$ and $\phi_{ss}$, and between $\phi_{bs}$ and $\phi_{sb}$, respectively.
  • Figure 4: Heatmaps of the fitted functions $f_{bb,sb}(\phi_{bb}, \phi_{sb})$ (left) and $f_{ss,bs}(\phi_{ss}, \phi_{bs})$ (right), which capture the bivariate effect between $\phi_{bb}$ and $\phi_{sb}$, and between $\phi_{ss}$ and $\phi_{bs}$, respectively.
  • Figure 5: Heatmaps of the fitted functions $f_{bb,bs}(\phi_{bb}, \phi_{bs})$ (left) and $f_{ss,sb}(\phi_{ss}, \phi_{sb})$ (right), which capture the bivariate effect between $\phi_{bb}$ and $\phi_{bs}$, and between $\phi_{ss}$ and $\phi_{sb}$, respectively
  • ...and 14 more figures