📈 economics

Competitive mediator games and urban CAV routing markets

This paper introduces competitive mediator games to model future autonomous routing and driving markets, proving that in generic anonymous congestion settings, market-share maximizing mediators lead to monopoly equilibria when one mediator is universally preferred, while also providing a comprehensive overview and mechanism design discussion for these emerging markets.

Original authors: Grzegorz Jamróz

Published 2026-08-11

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Grzegorz Jamróz

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world where your morning commute isn't a solo battle against traffic, but a game played by invisible referees. This is the realm of game theory, a branch of science that studies how people (or computers) make decisions when their success depends on what everyone else does. You might already know the concept of a Nash Equilibrium: a state where no one wants to change their move because everyone else is doing the same thing. Think of it like a crowded hallway where everyone is walking at the same speed; if you try to sprint, you just bump into people and slow down, so you stay put. But there's a catch: sometimes, if everyone could just agree to walk on the left side of the hall, everyone would get through faster. That's where mediated equilibria come in. Imagine a referee who whispers a secret instruction to each person ("You go left, you go right") that helps everyone move more efficiently than they could on their own.

Now, picture a future where self-driving cars are everywhere. Instead of you deciding which route to take, you download an app that drives you. But here's the twist: there isn't just one app. There are dozens of competing companies, each with its own "referee" trying to get you to use their service. This paper asks a big question: What happens when these referee apps compete against each other? Do they work together to make traffic flow perfectly? Do they fight so hard that the system breaks? Or does one giant referee eventually swallow up the whole market? The author is exploring this because, as we move toward a world run by algorithms, understanding these digital power struggles is crucial for keeping our cities moving smoothly and fairly.

The Great App Showdown: A Tale of Digital Referees

Welcome to the future of your commute. You hop into your car, but instead of grabbing the wheel, you tap an app. This app is your mediator—a digital referee that decides your route and drives the car for you. But you have a choice: you can drive yourself, or you can pick one of several competing apps. Let's call them "RouteRacer," "AutoPath," and "DriveEasy."

In this new world, these apps aren't just trying to get you from Point A to Point B; they are in a fierce battle for your subscription. Their goal? To get as many users as possible. The more users they have, the more successful they are. This is a competitive mediator game.

The Rules of the Game

Think of the city as a giant board game with two main routes: a fast highway and a scenic, slower road.

The Players: You (the driver) and the Apps (the mediators).
The Goal: You want to get to work with the least amount of "pain." For some, pain is just time. For others, being in a self-driving car is so relaxing that 15 minutes feels like 7.5 minutes. For skeptics, being in a robot car feels twice as long as driving themselves.
The Apps' Strategy: Each app has a secret playbook. They can tell their users to take the fast road, the slow road, or split them up randomly. They want to pick a strategy that makes their users happy enough to stay, while also making the traffic flow well enough to keep the users coming back.

The paper introduces a fascinating concept called the Competitive Mediator Equilibrium. This is the moment when the apps have found their perfect strategies, and you have found your perfect app. No app wants to change its routing plan, and no driver wants to switch apps.

The Big Discovery: The "Slight Advantage" Trap

Here is where the story gets a little scary, but also very logical. The author ran the math on this scenario, imagining a city full of drivers and two competing apps. They asked: What happens if one app is just a tiny bit better than the other?

Maybe "RouteRacer" has a slightly smoother interface, or maybe its drivers feel just a little more comfortable in its cars. In the world of these games, the author found that a tiny advantage can lead to a total monopoly.

Imagine a race where one runner has a shoe that is 1% lighter. In a normal race, that might not matter. But in this digital game, that 1% advantage allows the runner to change their strategy in a way that makes the other runner's strategy useless. The "better" app can use a clever, randomized strategy (like flipping a coin to decide who takes which road) that perfectly balances the traffic. Because of this, every single driver realizes that switching to the "worse" app would actually make their trip longer or more annoying.

The result? Everyone jumps to the "better" app. The second app gets zero customers. The market becomes a monopoly.

The paper proves this mathematically. If one mediator is weakly preferred by everyone (meaning no one thinks the other app is better) and strictly preferred by even a small group of people, that mediator can eventually take over the entire market. It's not because they are evil or colluding; it's just that their slight edge allows them to play the game so perfectly that the competition collapses.

Why This Matters (And Why It Might Be Okay)

You might think, "Oh no, a monopoly! That's bad!" And usually, in the real world, monopolies are bad for consumers because prices go up and quality goes down. But here is the twist: This isn't a monopoly of prices; it's a monopoly of quality.

In this specific market, the apps don't charge you money. Their "revenue" is simply having more users. So, if "RouteRacer" wants to keep its monopoly, it must keep being the best. If it stops improving, or if "AutoPath" invents a new feature that makes even a tiny fraction of users happier, the whole monopoly could shatter instantly.

The author suggests that this kind of competition might actually be a good thing. It forces the apps to constantly innovate and improve the "experience" of the ride (making it more relaxing, safer, or faster) to win over even a few extra users. It's a market where the only way to win is to be the absolute best.

What's Left to Figure Out?

The paper is a rigorous mathematical proof, not a simulation of real traffic yet. The author is very clear about what they know and what they don't.

They proved: In a simplified, theoretical world, a slight advantage leads to a monopoly.
They suspect: This might happen in the real world, but they haven't tested it on actual roads with real human drivers yet.
They wonder: What happens if the apps don't just compete but try to trick each other? What if the drivers are not perfect at math and make mistakes? What if the city has thousands of roads instead of just two?

The paper ends by saying that while we have a strong mathematical model for this, the real world is messy. We need to watch closely as these technologies roll out. If we let the market run wild, we might end up with one giant app controlling all our drives. But if that app is constantly fighting to be the best, maybe that's exactly what we need.

So, the next time you see a self-driving car, remember: it's not just a car. It's a player in a high-stakes game where a tiny advantage could change the entire city. And the winner of that game might just be the one that makes your commute feel the shortest, even if the clock says otherwise.

Technical Summary: Competitive Mediator Games and Urban CAV Routing Markets

1. Problem Statement

The paper addresses the emerging strategic landscape of Autonomous Routing and Driving (ARAD) markets, where multiple service providers (mediators) compete for users to manage their vehicle routing and driving. Unlike traditional pricing markets (e.g., petrol or retail), these markets involve complex welfare factors beyond price, such as travel time, user experience, and the value of time while autonomous.

The core problem is to model the competition between these strategic mediators who offer routing services. The author posits that current game-theoretic frameworks, such as Nash Equilibria and Correlated Equilibria, are insufficient because they do not account for:

Strategic Mediators: Entities that not only recommend actions but actively compete for market share and possess their own utility functions.
User Heterogeneity: Users have varying "discount factors" (relative value of time) depending on their trust in or preference for specific ARAD providers.
Market Structure: The potential for these markets to evolve into monopolies or oligopolies and the implications for social welfare and congestion.

The paper seeks to formalize a framework for Competitive Mediator Games (CMG) and analyze their equilibria, specifically within the context of future urban Collective Autonomous Vehicle (CAV) routing.

2. Methodology

The paper develops a general game-theoretic framework extending the concepts of Correlated Equilibria (CE) and Mediated Equilibria.

Game Formulation:
- Players: A set of users ( $N$ ) and a set of mediators ( $F$ ).
- Choices: Users choose between a set of actions (routes) $A$ or a set of mediators $F$ . Choosing a mediator delegates the action selection to that mediator.
- Utilities:
  - Users: Characterized by a set of disutility functions $\{u_i^f\}$ , where $u_i^0$ is the cost of independent driving and $u_i^f$ is the cost when mediated by $f$ . The cost is modeled as a product of travel time and a user-specific discount factor $\gamma_i^f$ (representing the value of time in an AV provided by $f$ ).
  - Mediators: Their utility functions $U_f$ depend on the joint actions of users and the users' choices. The paper focuses on share-maximizing mediators, where $U_f$ is proportional to the number of users choosing mediator $f$ .
- Stages:
  1. Mediators propose recommendation matrices (strategies) for any subset of users.
  2. Users select a mediator or an independent action.
  3. Actions are sampled based on the mediators' strategies, and payoffs are realized.
Equilibrium Concepts:
- Competitive Mediated Equilibrium (CME): Defined as a Nash equilibrium in the extended game where mediators are strategic players. It requires:
  - User Equilibrium: No user can improve their expected utility by unilaterally changing their choice of mediator or action, given the mediators' strategies.
  - Mediator Equilibrium: No mediator can improve its utility (market share) by unilaterally changing its recommendation strategy, given the strategies of other mediators and the resulting user equilibrium.
- Continuum of Players: The framework is extended to non-atomic games (infinitesimal users) using measure-theoretic formulations (following Mas-Colell) to handle large populations and anonymous congestion effects.
Specific Model: The theory is applied to Discounted Share-Maximizing Independent Routing Games. In this model, travel time is a function of congestion (flow), and mediators compete by optimizing routing strategies to attract users with specific discount factors.

3. Key Contributions

Formal Definition of Competitive Mediator Games: The paper introduces a rigorous definition of CMGs, generalizing Coarse Correlated Equilibria to settings with multiple strategic intermediaries competing for users.
Existence of User Equilibrium: The author proves that for any fixed set of mediator recommendation strategies, a user equilibrium exists (Proposition 6.7, Theorem 6.19).
Monopoly Result in Dominant Settings: The paper's primary theoretical result (Theorem 7.17) demonstrates that in a discounted share-maximizing routing game with two mediators and equivalent routes, if one mediator is strictly dominant (offering a weakly better discount factor to all users and strictly better to a non-negligible subset), the market converges to a monopoly. The dominant mediator can employ a randomized routing strategy that captures 100% of the market share, regardless of the competitor's strategy.
Stability of Monopoly: The paper shows (Proposition 7.19) that this monopoly equilibrium is robust; even if a large majority of users slightly prefer the non-dominant mediator, the dominant mediator can still secure the entire market if it has a "hard" advantage (e.g., a subset of users who strictly prefer it unconditionally) and utilizes randomization.

4. Results

Market Dynamics: In the specific context of ARAD providers maximizing market share, a slight advantage in service quality (modeled as a discount factor) can lead to a monopoly.
Inefficiency of Simple Routing: The paper notes that simple routing strategies (e.g., assigning all users to one route) are generally suboptimal for mediators trying to maintain market share against competitors.
Existence vs. Uniqueness: While user equilibria exist, the existence of a Competitive Mediator Equilibrium (CME) is more delicate. It relies on the uniqueness of the user equilibrium for a given set of mediator strategies. If multiple user equilibria exist, mediators may have conflicting preferences over which equilibrium forms, potentially preventing a CME from existing.
Welfare Implications: The paper argues that while the resulting monopoly might seem negative, the competition to achieve the "dominant" status incentivizes mediators to improve service quality (lowering discount factors), which could benefit users. However, the paper does not claim this mechanism is optimal for social welfare without further regulation.

5. Significance and Claims

The paper claims to provide the first general game-theoretic framework for markets with multiple strategic mediators competing for users, specifically tailored to the future of autonomous driving.

Theoretical Contribution: It bridges the gap between standard non-cooperative game theory (Nash) and mediated games, introducing the concept of equilibrium where the mediators themselves are competing agents.
Practical Insight: The results suggest that in fee-free, market-share-driven ARAD markets, natural competition may not lead to a diverse ecosystem of providers but rather to a monopoly if one provider gains a slight quality edge. This challenges the assumption that competition inherently prevents monopolies in algorithmic markets.
Policy Relevance: The paper highlights the need for mechanism design that considers these dynamics. It suggests that while a fee-free market-share model avoids predatory pricing, it may require additional regulatory interventions (e.g., incorporating social welfare into mediator objectives) to prevent undesirable monopolies.
Limitations: The author is modest about the scope, acknowledging that the existence of CME in general settings (especially with non-unique user equilibria) remains an open problem. The paper focuses on static, one-shot games and does not fully address dynamic, out-of-equilibrium learning processes or complex network topologies, though these are identified as future research directions.

In summary, the paper argues that algorithmic decision-making in routing markets introduces a new class of strategic interactions where "mediator competition" can paradoxically lead to market concentration, necessitating a re-evaluation of how these future markets should be regulated.