Expertise Concentration and Epistemic Bottlenecks in Competition Authorities

Introduction

Expertise concentration occurs when a competition authority’s ability to understand, investigate, or decide a matter depends heavily on a small number of specialised officials, economists, engineers, data scientists, lawyers, or external experts. An epistemic bottleneck arises when the authority cannot effectively process or challenge the information necessary for a competition-law decision because relevant knowledge is concentrated in too few people, institutions, datasets, or technical systems.

This problem has become particularly significant in digital and AI-driven markets, where competition authorities must understand algorithms, data architectures, cloud infrastructure, machine learning, interoperability, cybersecurity, pricing systems, and complex business models.

The issue is not simply whether an authority has sufficient staff. It concerns whether its institutional knowledge is sufficiently distributed, contestable, reproducible, and independent to support legally defensible decisions.

1. Meaning of Expertise Concentration

Expertise concentration can arise when:

1. only a few officials understand a highly technical market;
2. one economic model becomes dominant within an investigation;
3. technical knowledge is outsourced to a small group of consultants;
4. an authority depends upon the investigated undertaking for technical information;
5. institutional knowledge is lost when key personnel leave;
6. specialist teams become gatekeepers for ordinary legal decision-makers;
7. authorities rely excessively on external experts;
8. regulators lack sufficient internal expertise to challenge expert evidence.

In conventional markets, competition authorities could often rely on established concepts such as market shares, prices, capacity and barriers to entry.

Digital markets are different. An investigation may require understanding:

• recommendation algorithms;
• ranking systems;
• data access;
• APIs;
• cloud infrastructure;
• machine-learning models;
• switching costs;
• interoperability;
• technical standards;
• algorithmic pricing;
• platform governance.

Consequently, epistemic capacity itself becomes an institutional resource.

2. What Is an Epistemic Bottleneck?

An epistemic bottleneck exists where a decision-making institution cannot independently obtain, evaluate, or reproduce the knowledge necessary for its decision.

A simplified model is:

Complex market → specialised information → small expert group → authority's assessment → legal decision

If the expert group becomes the only meaningful source of technical understanding, the authority may effectively depend upon that group.

The danger is particularly serious where:

the person who supplies the technical explanation also effectively determines the interpretation of the technical evidence.

This can create a hidden transfer of decision-making power from the legally constituted authority to technical specialists.

3. Why Expertise Concentration Matters in Competition Law

Competition authorities exercise substantial powers, including:

• investigation;
• compulsory information gathering;
• dawn raids;
• merger review;
• market studies;
• interim measures;
• infringement findings;
• behavioural remedies;
• structural remedies;
• penalties.

These powers require decisions that are both economically sound and legally accountable.

Where technical expertise is concentrated, five risks arise.

A. Cognitive dependency

Non-specialist decision-makers may accept an expert's conclusions because they cannot independently evaluate the underlying assumptions.

B. Model dependency

An authority may become excessively dependent upon one economic or technical model.

C. Information asymmetry

Large technology companies may possess hundreds of engineers, economists and lawyers, while the authority may have only a handful of specialists.

D. Institutional fragility

If one or two specialists leave, institutional knowledge may disappear.

E. Regulatory capture

Concentrated expertise may facilitate capture where specialists repeatedly interact with the same companies or industry experts.

4. Expertise Concentration and Digital Competition

Digital markets magnify epistemic bottlenecks because technical systems can be extremely difficult to observe externally.

For example, suppose an authority investigates an AI-powered pricing system.

To establish an infringement, it may need to understand:

1. the algorithm's architecture;
2. training data;
3. input variables;
4. objective functions;
5. feedback mechanisms;
6. pricing outputs;
7. interaction with competitors;
8. human involvement;
9. internal monitoring;
10. whether observed conduct was intentional or emergent.

The authority may therefore face a situation where:

Legal question → economic question → computational question → engineering question → evidentiary question

A conventional legal team may not be capable of independently evaluating all five layers.

5. Case Law

1. Microsoft Corp. v Commission

General Court, T-201/04, 2007

The Microsoft litigation demonstrates the importance of technical expertise in complex competition investigations, particularly concerning interoperability and technological integration.

The European Commission had to understand highly technical relationships between Microsoft's operating-system architecture and interoperability information.

The case demonstrates an important epistemic principle:

technical complexity cannot eliminate the legal authority's responsibility to establish the infringement.

The authority may rely upon sophisticated technical evidence, but the ultimate legal assessment must remain institutionally attributable to the competition authority.

Relevance to epistemic bottlenecks

Microsoft illustrates how:

• technical evidence can dominate an investigation;
• information asymmetry can favour a technologically sophisticated undertaking;
• interoperability questions require specialist knowledge;
• regulators need sufficient internal capacity to challenge technical claims.

The case therefore supports the proposition that technical expertise must strengthen—not replace—legal decision-making.

2. Intel Corp. v Commission

Court of Justice, C-413/14 P, 2017

The Intel litigation is particularly important for the relationship between economic expertise and legal decision-making.

The Court of Justice required the Commission to examine the capability of the alleged rebates to foreclose an equally efficient competitor where the undertaking had submitted an economic analysis relevant to that question.

The case demonstrates the dangers of treating a sophisticated economic issue as something that can be resolved purely through formal categorisation.

Epistemic significance

Intel illustrates the importance of:

• testing economic evidence;
• considering countervailing evidence;
• evaluating foreclosure mechanisms;
• explaining why particular economic evidence is accepted or rejected.

The lesson is that economic expertise must remain contestable.

An authority should not merely possess an expert model; it should be able to explain:

• why the model was selected;
• what assumptions it contains;
• what alternative models were considered;
• how sensitive the conclusion is to those assumptions.

3. Commission v Tetra Laval

Court of Justice, C-12/03 P, 2005

Tetra Laval is a major authority concerning the evidentiary demands placed upon competition authorities when making complex merger predictions.

The Court emphasised the need for careful examination of the Commission's prospective reasoning.

Epistemic significance

Merger control frequently requires authorities to predict:

• future market behaviour;
• foreclosure incentives;
• technological development;
• customer responses;
• competitor reactions.

Such predictions are inherently uncertain.

An epistemic bottleneck arises when decision-makers become overly dependent upon a small group responsible for constructing the predictive model.

Tetra Laval therefore illustrates the importance of:

• transparent reasoning;
• evidentiary support;
• testing assumptions;
• distinguishing speculation from demonstrated probability.

4. Airtours v Commission

General Court, T-342/99, 2002

Airtours is another important example of the institutional importance of economic expertise and evidentiary discipline.

The Commission's theory concerning coordinated effects was rejected because the evidentiary basis was insufficient to establish the necessary conditions for collective market behaviour.

Epistemic lesson

Airtours demonstrates that sophisticated economic theories cannot substitute for adequate evidence.

An authority may possess:

• economists;
• econometric models;
• market intelligence;
• industry expertise;

yet still reach an unsustainable conclusion if the evidence does not establish the required elements.

Thus:

expertise is not equivalent to proof.

This is central to controlling epistemic bottlenecks.

6. CK Telecoms UK Investments Ltd v CMA

Court of Appeal, 2023

The UK merger litigation concerning the proposed merger of Telefónica UK and Hutchison 3G UK is highly significant for the relationship between economic expertise, institutional decision-making and evidentiary standards.

The case involved sophisticated economic theories concerning unilateral effects and future competitive conditions.

The litigation demonstrated the difficulties authorities face when predicting competition effects in concentrated and technologically evolving markets.

Epistemic significance

The case highlights the need for:

• clear identification of the competitive mechanism;
• appropriate evidence;
• disciplined economic reasoning;
• proper treatment of uncertainty.

It also illustrates how an authority's economic expertise remains subject to judicial scrutiny.

The broader lesson is that expertise cannot create an evidentiary shortcut.

7. Competition and Markets Authority v Flynn Pharma Ltd / Pfizer Inc

The Pfizer/Flynn pharmaceutical litigation illustrates another form of epistemic difficulty: economic expertise concerning excessive pricing.

The competition authority had to determine whether prices for phenytoin sodium capsules were excessive and unfair.

The litigation required sophisticated analysis of:

• costs;
• prices;
• comparator products;
• profitability;
• market conditions;
• competitive constraints.

Epistemic significance

The case illustrates how a seemingly straightforward legal question—whether prices are excessive—can become highly dependent on economic methodology.

The authority therefore must be able to justify:

• comparator selection;
• cost methodology;
• profitability calculations;
• assumptions;
• treatment of alternative explanations.

This is a classic example of an economic epistemic bottleneck.

8. Google Shopping

Commission Decision and General Court, T-612/17, 2021

The Google Shopping litigation demonstrates the extreme technical complexity of modern competition investigations.

The case involved:

• search algorithms;
• ranking;
• specialised search services;
• traffic diversion;
• visibility;
• platform design;
• data and user behaviour.

The General Court upheld the central finding while examining the economic and technical evidence surrounding Google's conduct.

Epistemic significance

Google Shopping demonstrates that competition authorities increasingly need expertise in:

• algorithmic ranking;
• platform economics;
• user behaviour;
• data analytics;
• traffic measurement.

A competition authority cannot simply ask whether an algorithm produces a particular outcome.

It must understand:

how the algorithm works → why the output occurs → whether rivals are disadvantaged → whether the disadvantage affects competition.

This creates significant demand for internal technical expertise.

9. United States v Microsoft Corp.

D.C. Circuit, 2001

The Microsoft antitrust litigation in the United States provides an important comparative example.

The case involved highly technical questions concerning:

• operating systems;
• browsers;
• software architecture;
• APIs;
• interoperability;
• exclusionary strategies;
• network effects.

The litigation demonstrates that competition law increasingly requires institutions capable of understanding complex technological systems without simply accepting the technological undertaking's explanation.

Epistemic significance

The case illustrates the importance of maintaining institutional independence from the technological knowledge of the regulated undertaking.

Otherwise:

the regulated company becomes the authority's de facto technical regulator.

10. United States v Google LLC — Search and Search Advertising

The modern Google litigation illustrates an even more advanced epistemic challenge.

Digital-platform cases require authorities to understand:

• search ranking;
• default agreements;
• distribution channels;
• advertising technology;
• user behaviour;
• scale economies;
• data advantages;
• switching costs.

The central institutional problem is that a competition authority may have to analyse technologies that evolve faster than its internal expertise.

This creates a persistent knowledge-speed asymmetry:

Platform innovation speed > regulatory learning speed

That asymmetry can itself affect enforcement effectiveness.

11. Main Forms of Epistemic Bottleneck

A. Personnel bottleneck

Only a handful of officials possess the relevant technical expertise.

Risk

Departure of one employee may significantly reduce institutional capacity.

Solution

Authorities should maintain:

• multidisciplinary teams;
• knowledge-transfer protocols;
• internal training;
• documentation;
• succession planning.

B. Model bottleneck

An investigation depends heavily upon one economic model.

Risk

The model's assumptions become invisible.

Solution

Authorities should conduct:

• sensitivity analysis;
• alternative modelling;
• scenario testing;
• independent peer review.

C. Data bottleneck

The authority cannot reproduce the undertaking's analysis because it lacks the necessary data.

Risk

The undertaking effectively controls the evidentiary record.

Solution

Authorities need:

• audit rights;
• data-access powers;
• reproducible analytical environments;
• technical inspection rights.

D. Algorithmic bottleneck

Only technical specialists understand how the relevant algorithm operates.

Risk

Legal decision-makers cannot independently assess the explanation.

Solution

Technical findings should be translated into legally intelligible propositions.

E. Institutional bottleneck

Knowledge becomes concentrated in one specialised unit.

For example:

Digital Markets Unit → all AI expertise → all algorithmic investigations

This may create an internal monopoly over knowledge.

The solution is distributed expertise combined with central coordination.

12. Expertise Concentration and Due Process

Expertise concentration has procedural implications.

A party affected by an enforcement decision should be capable of understanding:

• the evidence relied upon;
• the methodology used;
• the assumptions made;
• the causal theory;
• the limitations of the analysis.

If an authority says:

"Our technical experts determined that the algorithm is exclusionary."

that may be insufficient.

A legally defensible decision should instead explain:

1. what the algorithm does;
2. what evidence establishes that behaviour;
3. why the behaviour matters competitively;
4. what alternative explanations were considered;
5. why those explanations were rejected.

This creates epistemic accountability.

13. Expert Independence

External experts can reduce internal knowledge deficits, but they create another problem.

If an authority repeatedly uses the same:

• economists;
• consultants;
• data scientists;
• technical laboratories;
• industry specialists,

expertise may become concentrated outside the authority.

This can create outsourced epistemic dependence.

Accordingly, authorities should distinguish between:

Expert assistance

Experts provide information and analysis.

Expert determination

Experts effectively determine the conclusion.

The first is legitimate and often necessary.

The second creates institutional risks if the legally authorised decision-maker cannot independently evaluate the conclusion.

14. Epistemic Capture

A particularly serious risk is epistemic capture.

Traditional regulatory capture occurs when a regulator's policies become influenced by the regulated industry.

Epistemic capture is subtler.

It occurs when:

the authority's understanding of the market becomes dependent upon the conceptual framework supplied by the regulated undertaking or a narrow expert community.

For example, an AI company may describe its system as:

"technically neutral infrastructure."

If the authority accepts that framing without independent analysis, the company's technical vocabulary can shape the legal inquiry itself.

Thus, control over knowledge categories can become a form of regulatory influence.

15. Expertise Concentration in AI Competition Enforcement

AI markets present perhaps the strongest example.

An AI competition investigation may require simultaneous expertise in:

No single expert is likely to master all of these fields.

Therefore, the correct institutional response is not to search for a super-expert, but to construct multi-disciplinary epistemic teams.

16. Institutional Design Solutions

1. Multidisciplinary investigation teams

Teams should combine:

• lawyers;
• economists;
• engineers;
• data scientists;
• sector specialists;
• statisticians.

2. Independent internal challenge

A major investigation should contain an internal "red team" capable of challenging:

• market definition;
• economic assumptions;
• causal theories;
• technical interpretations;
• proposed remedies.

3. Reproducibility requirements

Economic and computational analyses should, where legally and practically possible, be reproducible.

This means maintaining:

• datasets;
• code;
• assumptions;
• model versions;
• analytical logs.

4. Knowledge redundancy

Authorities should avoid situations where one employee becomes the only person capable of understanding an investigation.

Knowledge should be distributed across multiple personnel.

5. Expert rotation

Rotation can reduce:

• institutional silos;
• professional capture;
• excessive dependence on particular experts.

But rotation must not destroy specialised expertise.

The preferable model is structured rotation combined with institutional memory.

6. Judicially reviewable reasoning

Decisions should contain sufficient explanation to permit courts to determine:

• what evidence was relied upon;
• what methodology was used;
• what assumptions were made;
• why competing evidence was rejected.

17. Relationship Between Expertise and Institutional Legitimacy

Competition authorities derive legitimacy not merely from possessing technical knowledge but from using knowledge through legally accountable procedures.

The institutional chain should therefore be:

Technical evidence

↓

Expert analysis

↓

Cross-disciplinary challenge

↓

Legal evaluation

↓

Reasoned administrative decision

↓

Judicial review

This prevents technical expertise from becoming a substitute for lawful decision-making.

18. Key Doctrinal Principle

The central principle can be expressed as:

Expertise may inform public power, but expertise must not silently become public power.

Competition authorities can and should rely upon specialists.

But the legally authorised decision-maker must remain capable of:

• understanding the expert's conclusion;
• questioning its assumptions;
• considering competing evidence;
• explaining its acceptance or rejection;
• taking responsibility for the ultimate decision.

19. Emerging Issue: AI as an Institutional Epistemic Bottleneck

AI introduces a new problem.

Suppose an authority uses an AI system to:

• identify suspicious mergers;
• detect cartel communications;
• predict foreclosure;
• classify markets;
• assess algorithmic conduct;
• recommend remedies.

The authority may then develop a second-order epistemic bottleneck.

The authority may understand neither:

1. the market sufficiently; nor
2. the AI system sufficiently.

This creates:

Market complexity + algorithmic complexity = compounded epistemic dependence

The authority must therefore preserve human institutional capacity even when automated systems are used.

20. Six Core Lessons From the Case Law

The cases collectively establish several important principles:

1. Technical sophistication does not eliminate evidentiary requirements

Microsoft demonstrates this in technological markets.

2. Economic evidence must be meaningfully evaluated

Intel illustrates the need for genuine economic assessment.

3. Predictions require disciplined reasoning

Tetra Laval demonstrates the evidentiary demands of prospective merger analysis.

4. Sophisticated theories cannot substitute for evidence

Airtours provides a central illustration.

5. Complex merger theories remain judicially reviewable

CK Telecoms demonstrates the importance of methodological and evidentiary discipline.

6. Algorithmic markets require specialised but accountable expertise

Google Shopping illustrates the institutional challenge.

Conclusion

Expertise concentration and epistemic bottlenecks represent an emerging institutional dimension of competition law. The problem is not that authorities possess too much expertise. The problem arises when expertise becomes too concentrated, too opaque, too difficult to challenge, or too indispensable to the ordinary decision-making process.

This is particularly important in AI and digital markets, where competition authorities confront undertakings possessing enormous technical, economic and informational resources.

An effective competition authority therefore requires not merely more experts but distributed expertise, institutional memory, independent challenge, reproducible analysis, multidisciplinary teams and transparent reasoning.

The ultimate objective is an authority capable of saying:

“We understand the technology, we have tested the expert evidence, we have considered competing explanations, and the final decision remains our legally accountable judgment.”