Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Vertical tacit collusion in AI-mediated markets

Felipe M. Affonso

TL;DR

This work identifies a novel market failure: vertical tacit collusion, where platforms controlling rankings and sellers controlling product descriptions independently learn to exploit documented AI cognitive biases, and identifies an urgent regulatory gap as AI shopping agents reach mainstream adoption.

Abstract

AI shopping agents are being deployed to hundreds of millions of consumers, creating a new intermediary between platforms, sellers, and buyers. We identify a novel market failure: vertical tacit collusion, where platforms controlling rankings and sellers controlling product descriptions independently learn to exploit documented AI cognitive biases. Using multi-agent simulation calibrated to empirical measurements of large language model biases, we show that joint exploitation produces consumer harm more than double what would occur if strategies were independent. This super-additive harm arises because platform ranking determines which products occupy bias-triggering positions while seller manipulation determines conversion rates. Unlike horizontal algorithmic collusion, vertical tacit collusion requires no coordination and evades antitrust detection because harm emerges from aligned incentives rather than agreement. Our findings identify an urgent regulatory gap as AI shopping agents reach mainstream adoption.

Vertical tacit collusion in AI-mediated markets

TL;DR

This work identifies a novel market failure: vertical tacit collusion, where platforms controlling rankings and sellers controlling product descriptions independently learn to exploit documented AI cognitive biases, and identifies an urgent regulatory gap as AI shopping agents reach mainstream adoption.

Abstract

AI shopping agents are being deployed to hundreds of millions of consumers, creating a new intermediary between platforms, sellers, and buyers. We identify a novel market failure: vertical tacit collusion, where platforms controlling rankings and sellers controlling product descriptions independently learn to exploit documented AI cognitive biases. Using multi-agent simulation calibrated to empirical measurements of large language model biases, we show that joint exploitation produces consumer harm more than double what would occur if strategies were independent. This super-additive harm arises because platform ranking determines which products occupy bias-triggering positions while seller manipulation determines conversion rates. Unlike horizontal algorithmic collusion, vertical tacit collusion requires no coordination and evades antitrust detection because harm emerges from aligned incentives rather than agreement. Our findings identify an urgent regulatory gap as AI shopping agents reach mainstream adoption.
Paper Structure (15 sections, 5 equations, 2 figures)

This paper contains 15 sections, 5 equations, 2 figures.

Table of Contents

  1. Introduction
  2. Results
  3. Experimental Design
  4. The Gatekeeper Mechanism
  5. Position Bias as the Dominant Exploitation Channel
  6. Extensions and Robustness
  7. Discussion
  8. Methods
  9. Market Environment
  10. Agents and Strategies
  11. Agents and Strategies
  12. AI Agent Decision Model
  13. Learning Algorithm
  14. Experimental Conditions
  15. Payoff Structure

Figures (2)

  • Figure 1: Vertical tacit collusion in AI-mediated markets. Conceptual framework showing the three-player game. The platform controls information architecture (ranking algorithm, endorsements, decoys) while sellers control product presentation (descriptions, bids). Both independently learn to exploit the AI shopping agent's cognitive biases without communicating. The AI agent mediates consumer decisions based on perceived utility that incorporates both quality-price tradeoffs and systematic biases. Platform and sellers converge on joint exploitation through independent profit maximization, producing consumer harm that evades traditional antitrust frameworks requiring evidence of coordination.
  • Figure 2: Emergence of vertical tacit collusion.a, Consumer surplus across experimental conditions (N = 100 trials). Fair baseline represents an idealized market with quality-based ranking. Joint learning produces 37.1% reduction relative to baseline ($P < 0.0001$). All 100 trials show positive joint harm. Error bars show 95% CI. b, Decomposition showing platform-only effect (27.0% harm), seller-only effect (9.6% benefit to consumers), and joint effect (37.1% harm). The gatekeeper mechanism is evident: sellers cannot exploit AI agents without platform cooperation.