Competition Law and Cognitive Network Effects and Market Power

1. Introduction

Cognitive network effects describe situations in which the value and competitive strength of a digital platform, ecosystem, or technology increase because users, businesses, developers, advertisers, or other participants generate data, information, behavioural feedback, knowledge, or algorithmic improvements through their participation.

Traditional network effects arise when a product becomes more valuable simply because more users join the network—for example, a communication platform. Cognitive network effects go further: each additional participant can generate information that improves the system itself, making the platform more attractive and potentially reinforcing its market position.

The competition-law concern is therefore not merely:

“More users make the platform more valuable.”

It is:

“More users generate data and learning that improve the platform, which attracts more users, generating still more data and learning.”

This can create a self-reinforcing cycle of scale → data → improved algorithms/services → greater attractiveness → more scale.

2. Meaning of Cognitive Network Effects

A cognitive network effect can be represented as:

More users
↓
More behavioural/data inputs
↓
Better algorithms, recommendations, prediction or matching
↓
Better service quality
↓
More users and commercial participants
↓
More data

This differs from a conventional network effect.

Traditional network effect

A telephone becomes more valuable because more people can be reached through it.

Cognitive network effect

A search engine, recommendation system, AI platform or digital marketplace can become more effective because more users generate:

• searches;
• clicks;
• purchases;
• ratings;
• reviews;
• location information;
• browsing behaviour;
• transaction histories;
• engagement data;
• developer interactions; and
• feedback used to train or refine algorithms.

The resulting competitive advantage may therefore be dynamic rather than static.

3. Cognitive Network Effects as a Source of Market Power

Competition law generally does not condemn network effects themselves. They become significant when they contribute to:

1. barriers to entry;
2. market concentration;
3. customer lock-in;
4. reduced multi-homing;
5. data advantages;
6. self-reinforcing dominance;
7. foreclosure of competitors;
8. raising rivals' costs;
9. exclusionary tying or bundling;
10. acquisition of nascent competitors; or
11. leveraging power between interconnected markets.

A particularly important issue is the possibility of a data feedback loop.

Data feedback loop

Incumbent platform

→ large user base

→ extensive data

→ superior prediction/targeting/recommendation

→ better user experience

→ increased user engagement

→ additional data

→ further improvement.

A smaller entrant may therefore face a problem beyond ordinary economies of scale.

The entrant may possess a technically competitive product but lack the quantity, diversity, velocity, or quality of data necessary to reproduce the incumbent's performance.

4. Relationship Between Network Effects and Dominance

Network effects do not automatically establish dominance.

Competition authorities ordinarily need to consider the relevant market and other factors, including:

• market shares;
• barriers to entry;
• switching costs;
• multi-homing;
• interoperability;
• access to data;
• economies of scale;
• financial resources;
• vertical integration;
• ecosystem effects;
• innovation;
• countervailing buyer power; and
• actual competitive constraints.

Thus:

Network effects ≠ dominance

but

Strong network effects + data advantage + switching costs + entry barriers
may substantially strengthen an undertaking's market power.

5. Cognitive Network Effects and Two-Sided Markets

Cognitive network effects are especially important in multi-sided platforms.

Examples include:

• search engines;
• online marketplaces;
• app stores;
• advertising platforms;
• social networks;
• payment systems;
• ride-hailing platforms;
• food-delivery platforms;
• cloud ecosystems;
• digital-content platforms.

A platform can have several interconnected participant groups.

For example:

Users → Platform ← Advertisers

Users generate behavioural information.

That information improves targeting.

Better targeting attracts advertisers.

Advertising revenue enables investment in the platform.

Improved services attract more users.

This creates a cross-side cognitive network effect.

6. Direct and Indirect Cognitive Network Effects

A. Direct cognitive network effects

The same side of the platform benefits from increased participation.

Example:

More users generate more interactions, which improve recommendation systems for other users.

B. Indirect cognitive network effects

Growth on one side improves the experience of another side.

Example:

More consumers → more transaction data → better marketplace analytics → more sellers → greater product variety → more consumers.

C. Cross-platform cognitive effects

A dominant ecosystem can use information generated in one service to improve another.

For example:

Search data → advertising optimisation → marketplace targeting → payment analytics.

Competition authorities may therefore examine whether data generated in one market strengthens market power in another.

7. Cognitive Network Effects and Barriers to Entry

One of the most important competition concerns is data-driven entry barriers.

Suppose an established platform has 90% of relevant transactions.

It receives:

• millions of searches;
• millions of purchases;
• millions of user reviews;
• millions of behavioural signals.

A new entrant begins with relatively little information.

Even if its algorithm is technically comparable, it may have insufficient data to provide equivalent predictive accuracy.

The incumbent therefore enjoys a data-scale advantage.

This may result in a barrier described as:

Economies of learning or learning-by-doing reinforced by network effects.

The competition-law question is whether that advantage arises from legitimate competition or is being protected or extended through exclusionary conduct.

8. Cognitive Network Effects and Switching Costs

Network effects can make switching particularly difficult.

Users may hesitate to leave because they would lose:

• contacts;
• reviews;
• transaction history;
• reputation;
• recommendations;
• personalised settings;
• accumulated loyalty benefits;
• social connections;
• stored information.

This creates data-based switching costs.

If switching is costly, an incumbent may retain users even where competing services offer better prices or technology.

9. Multi-Homing as a Competitive Constraint

An important counterweight to network effects is multi-homing.

Multi-homing means that a consumer or business uses several competing platforms simultaneously.

For example:

A seller may list products on several marketplaces.

An advertiser may use several advertising networks.

A developer may distribute applications through multiple ecosystems.

If multi-homing is easy, network effects may be less capable of producing durable market power.

If the dominant platform prevents or discourages multi-homing through:

• exclusivity;
• technical restrictions;
• contractual restrictions;
• discriminatory access;
• high switching costs; or
• interoperability limitations,

the network effect may become significantly more exclusionary.

10. Competition-Law Theories Applicable to Cognitive Network Effects

A. Abuse of Dominance

A dominant undertaking may potentially abuse its position through conduct designed to protect a data/network advantage.

Relevant theories include:

• exclusionary discrimination;
• refusal to supply;
• tying;
• bundling;
• self-preferencing;
• exclusivity;
• interoperability restrictions;
• exploitative data practices where legally relevant; and
• leveraging.

B. Foreclosure

A dominant platform could potentially use its network advantage to disadvantage competing platforms.

For example:

Dominant marketplace

→ restricts competitor access to essential behavioural information

→ competitor's recommendation system deteriorates

→ consumers migrate toward dominant marketplace.

The relevant issue is whether the conduct substantially forecloses competition rather than merely whether the incumbent possesses superior data.

C. Leveraging

Cognitive network effects can facilitate leveraging.

For example:

Search dominance

→ data advantage

→ advertising dominance

→ marketplace advantage.

Competition authorities may investigate whether market power in one market is being used to obtain or protect power in another.

11. Cognitive Network Effects and Merger Control

Cognitive network effects are particularly important in mergers involving digital platforms.

A merger may combine:

• two large user networks;
• complementary datasets;
• AI capabilities;
• advertising data;
• cloud infrastructure;
• consumer identity information.

Even where the parties' current market shares appear modest, the transaction may eliminate a nascent competitor capable of developing its own network and data advantage.

Merger analysis may therefore examine:

Horizontal effects

Will the transaction eliminate an existing competitor?

Conglomerate effects

Will combined datasets strengthen adjacent services?

Data effects

Will the merged firm obtain a unique or difficult-to-replicate dataset?

Innovation effects

Will the transaction reduce future innovation?

Network effects

Will combining networks make entry significantly more difficult?

12. Six Important Case Laws

1. United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001)

This is one of the foundational cases for understanding network effects and platform competition.

Microsoft possessed substantial power in the PC operating-system market. The court examined Microsoft's conduct concerning Internet Explorer, including restrictions affecting competing browsers.

Relevance

The case demonstrates how:

• network effects;
• applications ecosystems;
• platform compatibility;
• barriers to entry; and
• exclusionary conduct

can interact.

The larger the installed base, the greater the attractiveness of developing complementary applications for that platform. That can reinforce the incumbent's position.

Principle

Network effects may contribute to substantial barriers to entry, particularly where an incumbent uses exclusionary practices to protect an established platform.

2. United States v. Google LLC — Search Distribution Case (D.D.C. 2024)

The U.S. federal court examined Google's distribution arrangements involving search access points such as browsers and mobile devices.

The litigation concerned Google's position in general search and its agreements concerning default search placement.

Relevance to cognitive network effects

Search is a particularly strong example of a data-feedback system:

More searches

→ more behavioural information

→ improved search systems

→ greater user attraction

→ more searches.

Default placement can therefore have effects extending beyond immediate distribution.

Competition-law significance

The case illustrates how distribution arrangements can interact with:

• scale;
• data;
• network effects;
• default status; and
• barriers to expansion by rivals.

3. Google Search (Shopping), European Commission Decision, 2017

The European Commission found that Google had abused its dominant position in general search by systematically giving prominent placement to its own comparison-shopping service while demoting competing comparison-shopping services.

Relevance

Search platforms can accumulate enormous quantities of user interaction information.

A dominant search platform can potentially use its control over the search interface to influence traffic toward an affiliated service.

Competition issue

The important analytical connection is:

Search dominance + user data + control of interface + self-preferencing

may reinforce an ecosystem advantage.

The case is therefore highly relevant to understanding how control over a networked digital gateway can affect adjacent markets.

4. Google Android, European Commission Decision, 2018

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

• Google Search;
• Google Play;
• mobile-device manufacturers; and
• alternative search services.

The Commission concluded that certain practices restricted competition.

Cognitive-network relevance

Mobile ecosystems generate interconnected information concerning:

• application use;
• searches;
• consumer behaviour;
• device usage; and
• advertising interactions.

An ecosystem with a large installed base can consequently generate substantial informational advantages.

Principle

Platform ecosystems can produce reinforcing effects between interconnected products and services, making foreclosure analysis particularly important.

5. Facebook/Meta Data and Platform Cases — German Federal Cartel Office, 2019

The German Federal Cartel Office's proceedings concerning Facebook examined the relationship between Facebook's market position and its collection and combination of user data from different sources.

The authority considered Facebook's ability to combine data obtained from:

• Facebook;
• affiliated services; and
• third-party sources.

Relevance

This is particularly important for cognitive network effects because the competitive advantage of a platform can arise not merely from the number of users but from the breadth and combination of information generated by those users.

Competition-law significance

The case illustrates how data accumulation can be examined as part of a broader assessment of market power and exploitative/exclusionary conduct.

6. FTC v. Facebook, Inc. / Meta Platforms

The U.S. Federal Trade Commission's litigation concerning Facebook/Meta addressed alleged maintenance of monopoly power in personal social networking services.

The FTC's theory included Facebook's acquisition of Instagram and WhatsApp and the competitive significance of network effects.

Relevance

Social-network markets present particularly strong network effects because users value the service partly according to the number and quality of other users connected to it.

A large installed user base can therefore make entry difficult.

Cognitive dimension

The competitive significance becomes greater where the network also generates:

• behavioural information;
• social graphs;
• engagement data;
• advertising information; and
• personalization signals.

The case illustrates the importance of examining whether acquisitions reinforce an already powerful network ecosystem.

13. Additional Important Authorities

Several other cases are useful for understanding the broader legal framework.

7. Ohio v. American Express Co., 585 U.S. 529 (2018)

The U.S. Supreme Court treated the credit-card platform as a two-sided transaction platform and emphasized the need to consider both sides of the platform in assessing competitive effects.

Relevance

It demonstrates why platform markets cannot always be analysed by examining only one group of users.

8. European Commission — Microsoft (2004)

The Commission examined Microsoft's conduct concerning interoperability information and Windows Media Player.

Relevance

The case demonstrates the relationship between:

• platform dominance;
• interoperability;
• ecosystem control;
• foreclosure; and
• entry barriers.

These considerations remain relevant to modern cloud, AI and data ecosystems.

9. Google Android — European Commission (2018)

The Android decision also illustrates how contractual and ecosystem arrangements can reinforce network effects across interconnected digital services.

10. Intel Corp. v. European Commission

The Intel litigation demonstrates the importance of analysing exclusionary conduct by a dominant undertaking and the relationship between competitive foreclosure and market structure.

Although not a cognitive-network case specifically, it is useful when assessing whether conduct is capable of excluding competitors from a market where scale advantages already exist.

14. Cognitive Network Effects and Big Data

Big-data advantages can reinforce network effects through five characteristics:

1. Volume

More users produce more information.

2. Variety

Different types of information improve modelling.

3. Velocity

Real-time information can improve prediction.

4. Veracity

Large datasets can improve statistical reliability.

5. Value

The information can generate commercial or competitive advantages.

The resulting effect may be:

Network size → data accumulation → algorithmic improvement → greater network attractiveness.

This is sometimes described as a data-network feedback loop.

15. Cognitive Network Effects in AI Markets

AI systems provide an especially important modern example.

Consider an AI platform with:

• millions of users;
• extensive interaction data;
• feedback information;
• developer integrations;
• proprietary models;
• computing infrastructure.

More usage may generate information useful for:

• model evaluation;
• product improvement;
• personalization;
• safety testing;
• error detection;
• recommendation;
• workflow optimization.

Consequently:

Users → interactions → learning signals → better product → more users.

Competition authorities may therefore examine whether access to data, computing infrastructure, distribution, or users creates an entrenched competitive advantage.

16. Cognitive Network Effects and Essential Facilities

A particularly difficult question arises where a dataset or infrastructure becomes strategically important.

Potential examples include:

• payment transaction data;
• interoperability information;
• identity infrastructure;
• technical standards;
• cloud infrastructure;
• mapping information;
• search indexes;
• marketplace transaction data.

However, possession of an important dataset does not automatically make it an essential facility.

Competition law generally requires a careful analysis of:

• indispensability;
• feasibility of duplication;
• availability of alternatives;
• competitive foreclosure;
• objective justification; and
• applicable legal doctrine.

17. Cognitive Network Effects and Self-Preferencing

A platform may operate both:

1. the infrastructure through which users interact; and
2. a competing service on that infrastructure.

For example:

Marketplace operator

→ controls ranking algorithm

→ operates its own retail business

→ possesses marketplace data.

If it uses its platform position to preferentially promote its own service, the competition question becomes whether this conduct harms the competitive process.

The Google Shopping case provides an important framework for analysing this type of conduct.

18. Cognitive Network Effects and Algorithmic Pricing

Cognitive network effects can also arise in algorithmic pricing.

Suppose several platforms collect enormous amounts of:

• price data;
• demand information;
• consumer behaviour;
• competitor information.

Algorithms may become increasingly sophisticated as data accumulates.

Competition concerns may arise where:

• competitors use common pricing algorithms;
• algorithms facilitate coordination;
• platforms exchange competitively sensitive information;
• algorithmic systems independently produce parallel pricing; or
• a dominant platform uses data advantages to exclude rivals.

The legal analysis must distinguish independent algorithmic adaptation from conduct amounting to unlawful coordination or exclusion.

19. Cognitive Network Effects and Interoperability

Interoperability can reduce the strength of network effects.

For example:

Closed network

Users cannot easily communicate or transfer information outside the ecosystem.

→ high switching costs
→ strong network effect.

Interoperable network

Users can interact across competing services.

→ lower switching costs
→ greater multi-homing
→ weaker lock-in.

Thus interoperability may become a competition-law remedy in appropriate circumstances.

20. Possible Competition-Law Remedies

Where cognitive network effects contribute to anticompetitive conduct, possible remedies may include:

Structural remedies

• divestiture;
• separation of business units;
• restrictions on acquisitions.

Behavioural remedies

• non-discrimination;
• interoperability;
• data portability;
• access obligations;
• restrictions on self-preferencing;
• prohibition of exclusivity;
• transparency requirements.

Data-related remedies

• data portability;
• controlled data access;
• data-sharing obligations;
• restrictions on combining datasets;
• user consent requirements where relevant.

Merger remedies

• divestiture;
• licensing;
• interoperability commitments;
• firewall arrangements;
• data-access commitments.

The appropriate remedy depends on the particular theory of harm and the applicable jurisdiction.

21. Competition-Law Analytical Framework

A regulator examining cognitive network effects can proceed through the following sequence:

Step 1 — Define the relevant market

Determine:

• product/service market;
• geographic market;
• user groups;
• platform sides.

Step 2 — Identify the network effect

Ask:

Does additional participation increase the value of the platform?

Step 3 — Identify the cognitive component

Ask:

Does additional participation also generate information that improves the platform?

Step 4 — Examine feedback loops

Determine whether:

users → data → quality → users

creates self-reinforcement.

Step 5 — Measure market power

Consider:

• market shares;
• entry barriers;
• switching costs;
• data advantages;
• economies of scale;
• multi-homing;
• interoperability.

Step 6 — Identify exclusionary conduct

Investigate:

• tying;
• bundling;
• self-preferencing;
• exclusivity;
• discrimination;
• interoperability restrictions;
• refusal to provide access.

Step 7 — Assess competitive effects

Examine effects on:

• price;
• quality;
• innovation;
• privacy where legally relevant;
• consumer choice;
• market entry.

Step 8 — Consider efficiencies

A data/network advantage may result from legitimate investment and innovation.

Competition law should distinguish:

competition on the merits

from

conduct designed to exclude equally efficient or potentially viable competitors.

22. Key Distinction: Network Effect vs. Anticompetitive Conduct

This distinction is fundamental.

Legitimate network effect

A platform becomes more successful because consumers prefer it and participation creates genuine efficiencies.

Potential competition concern

A dominant platform deliberately prevents rivals from accessing customers, data, interoperability or distribution channels in order to preserve its network advantage.

Therefore, competition law does not normally punish success created by network effects.

The legal concern is the use of market power to exclude, foreclose, or disadvantage competition.

23. Emerging Issues

Cognitive network effects are likely to become increasingly important in:

• generative AI;
• foundation models;
• autonomous vehicles;
• digital health;
• fintech;
• cloud computing;
• smart-home ecosystems;
• digital advertising;
• e-commerce;
• app stores;
• search engines;
• social media;
• data marketplaces;
• digital identity systems;
• robotics; and
• industrial IoT.

The most significant future question may be whether data and learning advantages can become self-reinforcing barriers to entry even where traditional market-share measurements initially underestimate competitive power.

24. Conclusion

Cognitive network effects represent an advanced form of network-based market power in which users do not merely increase the value of a platform—they also generate information that can improve the platform itself.

The resulting cycle can be expressed as:

More users → more data → better algorithms/services → greater attractiveness → more users.

From a competition-law perspective, this mechanism can contribute to:

• durable market power;
• entry barriers;
• switching costs;
• reduced multi-homing;
• ecosystem expansion;
• data advantages;
• foreclosure; and
• strategic leveraging.

The leading authorities—including Microsoft, Google Search (Shopping), Google Android, the Facebook/Meta proceedings, American Express, and Microsoft interoperability litigation—provide useful legal foundations for analysing these issues, although not all of them involve the modern concept of “cognitive network effects” as such.

The central analytical question is therefore not whether a firm benefits from network effects, but whether network and data feedback mechanisms have created substantial market power and whether that power is being maintained or extended through conduct that harms the competitive process.