Competition Law and Intelligent Reflexive Ecosystems and Dominance

1. Introduction

An intelligent reflexive ecosystem is a digital, technological, or commercial ecosystem in which the system continuously observes market behaviour, learns from that behaviour, changes its own operation, and thereby influences the behaviour that it subsequently observes.

Examples include:

• AI recommendation ecosystems;
• intelligent procurement platforms;
• algorithmic pricing systems;
• smart logistics networks;
• app stores and operating systems;
• digital advertising ecosystems;
• AI-powered marketplaces;
• cloud-service ecosystems;
• smart manufacturing platforms;
• financial and payment ecosystems;
• data-driven search and ranking systems.

The competition-law problem arises when a dominant undertaking controls the ecosystem's data, algorithms, interfaces, standards, infrastructure, distribution channels, or feedback loops, allowing it to reinforce its own market position.

The basic concern can therefore be represented as:

Data → Prediction → Recommendation/Decision → User Behaviour → New Data → Improved Prediction → Greater Dependence → Stronger Market Power

This is a reflexive competitive feedback loop.

Competition law does not prohibit an ecosystem merely because it is intelligent, adaptive, or highly successful. The legal issue is whether its architecture or conduct is being used to exclude rivals, foreclose market access, exploit users or suppliers, extend dominance into neighbouring markets, or facilitate anticompetitive coordination.

2. Meaning of an Intelligent Reflexive Ecosystem

A conventional digital platform may simply provide a service.

An intelligent reflexive ecosystem goes further. It:

1. collects information about participants;
2. processes the information through algorithms or AI;
3. predicts behaviour;
4. modifies rankings, prices, recommendations or access;
5. causes users or businesses to respond;
6. collects the resulting information;
7. feeds that information back into the system.

Example

Suppose a dominant e-commerce platform controls:

• seller data;
• customer-search data;
• recommendation algorithms;
• advertising;
• logistics;
• payment;
• ranking;
• loyalty programmes.

The platform can observe that consumers increasingly purchase Product A.

Its algorithm then promotes Product A.

The increased visibility produces more purchases.

Those purchases create additional data.

The additional data makes the recommendation system apparently more accurate.

The platform then promotes its own competing Product A more aggressively.

The resulting cycle can create a self-reinforcing competitive advantage.

The competition concern is not simply "AI dominance." It is the possibility that the undertaking has transformed informational advantages into structural foreclosure.

3. Legal Framework

A. Relevant Market

The first question is normally:

What market is affected by the conduct?

Possible markets include:

• online marketplace services;
• search services;
• digital advertising;
• app distribution;
• mobile operating systems;
• cloud services;
• payment services;
• logistics;
• data-processing services;
• AI services;
• enterprise software;
• procurement platforms.

Traditional market-definition techniques can become difficult because many digital services have:

• zero monetary prices;
• multi-sided users;
• rapidly changing technology;
• network effects;
• data-driven quality competition.

Therefore, competition authorities may need to examine:

• user behaviour;
• switching costs;
• multi-homing;
• data advantages;
• interoperability;
• ecosystem dependencies;
• indirect network effects;
• algorithmic barriers to entry.

4. Dominance in Reflexive Ecosystems

Dominance may arise from several mutually reinforcing advantages.

4.1 Data advantage

A dominant platform may possess a volume and variety of data unavailable to competitors.

4.2 Algorithmic advantage

More data may produce better predictions.

Better predictions may attract more users.

More users produce more data.

This creates a data-network feedback loop.

4.3 Network effects

The value of a platform may increase as more:

• consumers;
• sellers;
• advertisers;
• developers;
• suppliers

join the ecosystem.

4.4 Switching costs

Users may become dependent on:

• stored data;
• purchase histories;
• loyalty benefits;
• APIs;
• proprietary formats;
• cloud infrastructure;
• payment systems;
• applications.

4.5 Ecosystem lock-in

A company may not dominate every individual market but may possess an ecosystem position that makes entry into neighbouring markets difficult.

5. Reflexive Dominance Versus Ordinary Dominance

Traditional dominance generally concerns the ability of an undertaking to behave independently of competitors, customers or suppliers.

Reflexive dominance has an additional characteristic:

The dominant undertaking's own conduct can continuously modify the competitive environment that subsequently reinforces its dominance.

For example:

Dominant platform

↓

Collects transaction data

↓

Improves AI recommendation

↓

Improves ranking accuracy

↓

Attracts more users

↓

Generates more transactions

↓

Generates more data

↓

Strengthens AI advantage

This makes dominance potentially dynamic rather than static.

6. Principal Competition Concerns

A. Self-preferencing

A dominant platform may use its intelligence system to favour:

• its own products;
• affiliated services;
• preferred suppliers;
• proprietary applications;
• internal logistics;
• its own advertising products.

The critical question is whether the platform is using control over an important ecosystem layer to disadvantage independent competitors.

B. Algorithmic exclusion

Algorithms may determine:

• search rankings;
• seller visibility;
• advertising placement;
• access to customers;
• procurement opportunities;
• pricing;
• recommendations.

An apparently neutral algorithm may nevertheless produce exclusionary effects.

The legal analysis should distinguish between:

1. legitimate algorithmic optimisation;
2. accidental discriminatory effects;
3. intentional exclusion;
4. structural self-preferencing;
5. exploitation of sensitive competitor data.

C. Data leveraging

A dominant undertaking may obtain commercially sensitive information from businesses operating on its platform.

It may then use that information to compete against those businesses.

For example:

Independent seller → supplies transaction data → platform analyses demand → platform launches competing product.

The competition issue becomes particularly serious where the platform acts simultaneously as:

• infrastructure provider;
• data collector;
• marketplace operator;
• competitor.

7. Feedback-Loop Foreclosure

The most distinctive issue in intelligent reflexive ecosystems is feedback-loop foreclosure.

A dominant platform can potentially create a loop such as:

More users → more data → better AI → better service → more users.

This is not automatically unlawful.

The concern arises when the undertaking artificially strengthens the loop through exclusionary conduct, such as:

• restricting interoperability;
• denying access to essential data;
• degrading competing services;
• tying products;
• manipulating rankings;
• self-preferencing;
• imposing exclusivity;
• preventing multi-homing;
• restricting portability.

The resulting barrier to entry may be significantly higher than a conventional price-based barrier.

8. Six Important Case Laws

1. Google Search (Shopping) — European Commission / General Court

Facts

Google was found to have systematically positioned and displayed its own comparison-shopping service more favourably in its general search results while competing comparison-shopping services were subject to Google's generic ranking mechanisms.

Competition principle

The case is highly relevant to reflexive ecosystems because Google controlled an important information and ranking infrastructure while simultaneously participating in the downstream comparison-shopping market.

The competition concern was therefore not simply Google's size.

It involved the use of a dominant upstream service to favour its own downstream offering.

Relevance to intelligent reflexive ecosystems

A recommendation or ranking system can become a competitive bottleneck.

Where the ecosystem determines:

• visibility;
• ranking;
• discoverability;
• traffic;

control over the algorithm can substantially influence competitive conditions.

Principle

A dominant platform's control over an important algorithmic access point may create competition concerns where that infrastructure is used to favour its own downstream service.

2. Google Android — European Commission / General Court

Facts

The European Commission examined Google's conduct concerning Android, including arrangements involving:

• Google Search;
• Google Chrome;
• Google Play Store;
• Android device manufacturers;
• mobile operating systems.

The Commission identified several practices that it considered capable of reinforcing Google's position in search and related markets.

Competition principle

The case demonstrates how several complementary services can form an ecosystem in which dominance in one layer reinforces dominance in another.

The ecosystem can operate through:

• defaults;
• contractual restrictions;
• application distribution;
• operating-system control;
• network effects.

Relevance

An intelligent ecosystem can similarly use:

operating system → data → applications → user behaviour → additional data

to reinforce market power.

Principle

Competition analysis may need to examine the interaction between several ecosystem layers rather than examining each product in complete isolation.

3. Amazon Marketplace Investigation — European Commission

Facts

The European Commission investigated Amazon's use of non-public marketplace seller data.

Amazon operated simultaneously as:

• marketplace intermediary; and
• retailer competing with marketplace sellers.

The concern was that Amazon could obtain commercially sensitive information generated by independent sellers and potentially use it in its own retail activities.

Competition principle

The case illustrates the danger created when an intermediary controls the information infrastructure through which its competitors operate.

Relevance to reflexive ecosystems

This is particularly important for AI ecosystems.

Imagine:

seller data → AI analysis → demand prediction → Amazon's product selection → better sales → additional seller/customer data.

The platform's intermediary position can therefore become an informational competitive advantage.

Principle

Control over a marketplace can create competition concerns where the platform simultaneously competes with the businesses whose commercially sensitive information it controls.

4. Meta Platforms / Facebook Data-Related Conduct — German Competition Authority

Facts

Germany's Bundeskartellamt examined Facebook's combination of user data obtained from Facebook with data obtained from other Meta services and external sources.

The case concerned the relationship between:

• market power;
• data collection;
• data combination;
• user choice;
• privacy-related conditions.

Competition principle

The case demonstrated that competition law can consider the relationship between market power and extensive data advantages.

Relevance

Reflexive ecosystems depend heavily upon data feedback.

If a dominant undertaking can combine information across multiple services, it may obtain a broader behavioural picture than competitors.

That may strengthen:

• targeting;
• recommendations;
• prediction;
• personalisation;
• advertising;
• product development.

Principle

Data aggregation can contribute to market power and may become particularly important where a dominant ecosystem uses information obtained from multiple services to reinforce its competitive position.

5. United Brands v Commission — Court of Justice of the European Union

Facts

United Brands was found to hold a dominant position in the relevant banana market and was held responsible for several forms of abusive conduct.

Competition principle

The case is a foundational authority on dominant position and abuse.

It established the classic understanding that dominance involves economic strength enabling an undertaking to behave to an appreciable extent independently of competitors, customers and consumers.

Relevance to reflexive ecosystems

Although United Brands predates digital technology, its dominance principle remains useful.

An intelligent ecosystem can possess economic strength through:

• data;
• algorithms;
• infrastructure;
• network effects;
• switching costs;
• ecosystem integration.

The technological mechanism changes, but the underlying competition-law question remains:

Does the undertaking possess sufficient market power to behave independently of competitive constraints?

6. Microsoft — European Commission / General Court

Facts

Microsoft faced European competition proceedings concerning practices involving its operating system and related software markets, including interoperability and tying-related issues.

Competition principle

The Microsoft litigation demonstrates the importance of interoperability where a dominant technological platform controls an important infrastructure layer.

Relevance

Modern intelligent ecosystems can similarly control:

• APIs;
• data formats;
• interoperability protocols;
• cloud interfaces;
• AI models;
• operating systems;
• developer access.

A dominant ecosystem may make competitors dependent upon infrastructure controlled by the incumbent.

Principle

Denial or restriction of interoperability can become particularly significant where competitors require access to a dominant technological platform to compete effectively.

9. Additional Relevant Case Laws

7. Intel v Commission

The Intel litigation is important for understanding exclusionary conduct involving rebates and the assessment of competitive effects.

Relevance

In an intelligent ecosystem, loyalty incentives can be combined with:

• algorithmic ranking;
• preferential access;
• platform rebates;
• exclusive arrangements.

The competitive assessment should therefore consider whether incentives reinforce foreclosure.

8. Bronner v Mediaprint

This case is important to the essential-facilities/refusal-to-deal doctrine.

Relevance

An intelligent ecosystem may become commercially indispensable where competitors cannot realistically reproduce:

• the data infrastructure;
• interoperability layer;
• network;
• technical interface;
• distribution infrastructure.

However, dominance alone does not automatically create a general duty to provide access.

9. Slovak Telekom

The case concerns exclusionary conduct involving access and margin-squeeze principles in telecommunications.

Relevance

It demonstrates how control over an upstream infrastructure layer can affect competition downstream.

For intelligent ecosystems, the same logic can arise where a platform controls:

infrastructure → access conditions → downstream competition.

10. Application to China

For a China-focused competition-law analysis, intelligent reflexive ecosystems can principally be examined under the Anti-Monopoly Law of the People's Republic of China (AML) and the rules and enforcement practice concerning the digital economy.

The principal provisions include:

• Article 6 — abuse of dominant market position;
• Article 17 — prohibited forms of abuse by dominant undertakings;
• Article 18 — factors for determining dominant market position;
• Article 19 — presumptions of dominance;
• provisions concerning platform-economy concentrations and monopoly agreements;
• the Anti-Monopoly Guidelines for the Platform Economy.

The Chinese framework is particularly relevant to algorithmic ecosystems because platform competition may involve:

• data;
• algorithms;
• technology;
• capital;
• platform rules;
• network effects;
• multi-sided markets.

11. Chinese Platform Cases Relevant to Reflexive Ecosystems

A. Alibaba — Abuse of Dominance

Alibaba's "choose one from two" enforcement action is particularly relevant to ecosystem dominance.

The conduct involved restrictions imposed upon merchants concerning operation across competing platforms.

Significance

Exclusivity can prevent merchants from multi-homing.

That matters enormously in a reflexive ecosystem because:

fewer competing platforms → more transactions on incumbent → more data → stronger algorithms → greater attractiveness → even more transactions.

Thus, exclusivity can reinforce a data-network feedback loop.

B. Meituan — Abuse of Dominance

The Meituan case concerned the use of market power in platform services and exclusive arrangements.

Relevance

Platform exclusivity can affect:

• merchant choice;
• rival platforms;
• consumer access;
• data accumulation.

In an intelligent ecosystem, preventing multi-homing can make the incumbent's data advantage self-reinforcing.

C. Didi — Platform and Data-Based Competition Issues

The Didi enforcement environment illustrates the importance of:

• platform control;
• data;
• algorithms;
• network effects;
• digital-market regulation.

Relevance

Ride-hailing platforms generate enormous quantities of:

• location information;
• demand data;
• supply information;
• pricing information;
• consumer behaviour.

These data can improve algorithmic matching and forecasting, potentially strengthening network effects.

12. Self-Preferencing in an Intelligent Ecosystem

Consider an AI marketplace.

The platform operates:

1. marketplace;
2. AI recommendation engine;
3. payment system;
4. logistics;
5. private-label products.

Suppose the algorithm repeatedly recommends the platform's own products.

The competitive cycle becomes:

Own product receives better recommendation
↓
Own product receives more sales
↓
Platform obtains more transaction data
↓
AI becomes better at predicting demand for own product
↓
Recommendation advantage increases
↓
Rival visibility declines

This can potentially create algorithmic self-reinforcement.

Competition authorities should therefore examine not only the immediate ranking decision but the long-term feedback effects.

13. Algorithmic Discrimination

An intelligent ecosystem may discriminate against rivals through:

• ranking degradation;
• reduced visibility;
• API throttling;
• slower integration;
• higher commissions;
• discriminatory search results;
• exclusion from recommendations;
• inferior interoperability.

The important question is whether the differentiation reflects legitimate technical or commercial criteria or whether it is being used as an exclusionary instrument.

14. Network Effects and Tipping

Reflexive ecosystems are particularly vulnerable to market tipping.

A market may move from:

several competing platforms

to:

one dominant ecosystem.

The mechanism may involve:

Users

↓

Data

↓

AI improvement

↓

Better service

↓

More users

↓

More data

This creates a positive-feedback mechanism.

Once tipping occurs, entry may become substantially harder because a new entrant must simultaneously overcome:

• data disadvantage;
• network effects;
• switching costs;
• established user relationships;
• ecosystem integration.

15. Interoperability as a Competition Remedy

Where ecosystem dominance is reinforced by technical barriers, possible competition remedies may include:

1. API access

Competitors may be given non-discriminatory technical access.

2. Data portability

Users may be allowed to transfer their data to competing services.

3. Interoperability

Different platforms may be required to communicate using appropriate technical standards.

4. Non-discrimination

The dominant platform may be prohibited from systematically disadvantaging competing services.

5. Separation of functions

In particularly serious circumstances, structural or behavioural separation may be considered.

16. Data Portability and the Reflexive Loop

Data portability can weaken the feedback loop.

Without portability:

User → incumbent → data accumulation → improved service → user retention.

With portability:

User → incumbent → transferable data → competing platform → alternative service.

Portability therefore potentially lowers:

• switching costs;
• data-entry barriers;
• network-effect barriers.

However, compulsory data access must also consider:

• privacy;
• cybersecurity;
• intellectual property;
• confidential information;
• proportionality.

17. Algorithmic Transparency

Competition law does not necessarily require a dominant company to disclose its complete algorithm.

Excessive disclosure could create:

• trade-secret problems;
• gaming;
• cybersecurity risks;
• reduced innovation.

Instead, competition authorities may investigate:

• ranking criteria;
• discriminatory outcomes;
• internal instructions;
• algorithmic changes;
• A/B testing;
• treatment of competitors;
• data use;
• communications between business and algorithmic teams.

18. Intelligent Procurement Ecosystems

The concept is particularly important in procurement.

Suppose an AI procurement platform controls:

• supplier qualification;
• tender invitations;
• pricing recommendations;
• supplier ranking;
• contract allocation.

If the platform systematically excludes certain suppliers, the AI may become a gatekeeper.

The competition concern can arise through:

• discriminatory access;
• exclusive procurement;
• algorithmic allocation;
• preferential treatment;
• information asymmetry;
• coordination among suppliers.

19. Algorithmic Collusion

Reflexive systems also create a different risk:

competitors may use algorithms that continuously observe and respond to one another.

Suppose competing firms employ pricing algorithms that:

• observe competitors' prices;
• react immediately;
• learn from previous interactions;
• optimise long-term profits.

The market may experience sustained high prices without a traditional explicit agreement.

Competition law must distinguish between:

Legitimate independent adaptation

and

Concerted or coordinated conduct.

The fact that algorithms independently reach similar prices does not, by itself, establish an unlawful agreement.

20. Consumer Welfare and Quality Competition

Reflexive ecosystems can produce significant benefits:

• better recommendations;
• lower search costs;
• personalised services;
• improved logistics;
• lower transaction costs;
• improved forecasting;
• fraud detection;
• better resource allocation.

Therefore, competition law should not treat algorithmic adaptation itself as anticompetitive.

The key distinction is:

Innovation-driven feedback versus exclusion-driven feedback.

21. Enforcement Test

A useful analytical framework is:

Step 1 — Identify the ecosystem

What services, platforms and technologies are interconnected?

Step 2 — Identify the bottleneck

What does the dominant undertaking control?

• data?
• API?
• operating system?
• algorithm?
• marketplace?
• payment system?
• cloud?
• logistics?

Step 3 — Determine dominance

Examine:

• market share;
• network effects;
• switching costs;
• multi-homing;
• data advantages;
• entry barriers.

Step 4 — Identify the feedback loop

How does conduct create:

advantage → behavioural response → additional data → stronger advantage?

Step 5 — Identify exclusion

Determine whether rivals suffer:

• foreclosure;
• reduced visibility;
• discriminatory access;
• higher costs;
• data disadvantage;
• interoperability restrictions.

Step 6 — Examine objective justification

Consider:

• security;
• privacy;
• efficiency;
• technical necessity;
• consumer protection;
• innovation.

Step 7 — Assess effects

Consider:

• actual foreclosure;
• potential foreclosure;
• consumer harm;
• reduced innovation;
• reduced choice;
• higher prices;
• lower quality;
• reduced market entry.

22. Compact Flowchart

Intelligent Ecosystem

↓

Data Collection

↓

Algorithmic Learning

↓

Prediction / Ranking / Pricing

↓

User or Supplier Behaviour

↓

Additional Data

↓

Reinforced Competitive Advantage

↓

Does the undertaking possess dominance?

→ No: ordinary competition/innovation analysis

→ Yes: examine conduct

↓

Self-preferencing / tying / exclusivity / discrimination / refusal of access / data leveraging

↓

Exclusionary or exploitative effects?

↓

Objective justification?

↓

Competitive assessment under applicable competition law

23. Key Legal Principles from the Case Law

24. Conclusion

Intelligent reflexive ecosystems create a distinctive competition-law problem because market power can become self-reinforcing.

The central mechanism is:

Data → Intelligence → Behavioural influence → More activity → More data → Greater intelligence → Greater market power.

The existence of this loop is not itself unlawful. Intelligent recommendations, adaptive pricing, personalisation and ecosystem integration may generate substantial efficiencies.

The competition-law concern arises where a dominant undertaking uses control over the loop to:

• exclude competitors;
• discriminate against rival suppliers;
• self-preference;
• restrict multi-homing;
• exploit competitor data;
• deny interoperability;
• tie complementary products;
• impose exclusionary conditions;
• leverage dominance into adjacent markets.

The most important conceptual shift is therefore from examining individual conduct in isolation to examining the architecture of the ecosystem and the feedback mechanisms through which conduct changes future competitive conditions.

In China, the Anti-Monopoly Law and Platform Economy Guidelines provide an especially relevant framework for analysing such issues through the combined concepts of dominance, platform rules, data, algorithms, network effects, exclusivity and digital-market foreclosure.