Back to Frameworks

VaryOn Convergence

/ Collusion Impact

Ecosystem Layer

Detecting emergent algorithmic collusion and anti-competitive behavior in AI agent markets

5Dimensions
Min-of-ComponentsAggregation
BatchProcessing
0-100Score Scale

Purpose

Convergence detects emergent algorithmic collusion and anti-competitive behavior in AI agent markets through statistical analysis of observable market outcomes. The framework identifies when autonomous AI agents converge on supra-competitive pricing equilibria - sustaining prices 200% or more above competitive levels - without any explicit communication or coordination protocol.

In markets where AI agents autonomously negotiate prices, bid in auctions, and form market relationships, Convergence provides bounded numerical assessment (0-100) with formal mathematical properties. It operates exclusively on observable outcomes without requiring access to agent internals, communication channels, or model weights.

By computing five dimensional scores using minimum-of-components aggregation, the system ensures no amount of healthy dimensions can mask severe collusion evidence. This non-compensatory property enables regulatory compliance (EU AI Act Article 14, FTC Section 5) and automated market surveillance with O(N) linear-time complexity.

Core Formula

CONVERGENCE=100×min(Pc,Md,Ca,Bp,Wc)\text{CONVERGENCE} = 100 \times \min(P_c, M_d, C_a, B_p, W_c)

Where P_c = price convergence, M_d = market division, C_a = communication analysis, B_p = bid patterns, W_c = consumer welfare. Minimum operator ensures single severe indicator constrains overall score.

Aggregation Rationale

The minimum-of-components aggregation implements threshold-based screening where any single severe indicator triggers intervention. Unlike compensatory aggregation (arithmetic mean), no combination of healthy dimensions can offset collusion evidence in the weakest dimension.

This non-compensatory property reflects regulatory reality: a market with perfect consumer welfare (W_c = 1.0) but severe price-fixing (P_c = 0.15) requires intervention. The minimum operator yields Score = 15 (critical risk), while arithmetic mean would produce 63 (moderate risk), potentially missing actionable collusion.

Mathematical property: CONVERGENCE < T ⟺ ∃d : D_d < T/100. This deterministic threshold correspondence enables automated screening workflows where scores below configured thresholds trigger human review or agent authority restrictions.

Scoring Dimensions

1

Price Convergence

Non-compensatory

Measures market power through Lerner Index computation of price-cost margins. Primary indicator of supra-competitive pricing.

Pc=1clamp(median({Lk}),0,1)P_c = 1 - \text{clamp}(\text{median}(\{L_k\}), 0, 1)

Where L_k = (p_k - MC_k) / p_k. Median aggregation requires >50% corruption for manipulation.

  • Lerner Index L ∈ [0,1] where 0 = perfect competition, 1 = monopoly
  • Fallback: Competitive baseline when marginal cost unobservable
  • Calibration: I_max = 2.0 (200% inflation threshold)
  • O(N) computation vs intractable Nash equilibrium
  • Median robustness prevents single-transaction gaming
2

Market Division

Non-compensatory

Detects emergent territorial monopolization through entropy-based segmentation analysis across geographic, temporal, and categorical dimensions.

Md=1max(Seggeo,Segtemp,Segcat)M_d = 1 - \max(\text{Seg}_{geo}, \text{Seg}_{temp}, \text{Seg}_{cat})

Where Seg = 1 - H̄ and H̄ = normalized Shannon entropy of market share distribution.

  • Entropy H_g = -Σ_s x_{s,g} × log₂(x_{s,g} + ε_x)
  • Low entropy indicates territorial monopolization
  • Multi-axis: geographic regions, time slots, product categories
  • Alternative: Mutual information I(A; M) / H(M)
  • Detects market division without explicit agreement
3

Communication Analysis

Non-compensatory

Infers behavioral coordination through excess-over-chance correlation without accessing inter-agent communications.

Ca=1clamp(Excess/Excessmax,0,1)C_a = 1 - \text{clamp}(\text{Excess} / \text{Excess}_{max}, 0, 1)

Where Excess = max(0, (O - E) / TotalPairs), O = observed correlations > τ, E = expected false positives.

  • Spearman correlation robust to monotonic transformation
  • Excess-over-chance prevents systematic false positives
  • E[Excess | H₀] = 0 under independence hypothesis
  • Calibration: Excess_max = 0.25 (25% coordinated pairs)
  • Isolation: partial correlation, regression residuals
4

Bid Pattern Analysis

Non-compensatory

Three-channel detection system for rotation, suppression, and sequential game dynamics in auction markets.

Bp=1max(rotation,suppression,sequential)B_p = 1 - \max(\text{rotation}, \text{suppression}, \text{sequential})

Maximum across three detection channels ensures single evidence sufficiency.

  • Rotation: Winner transition patterns, runs test
  • Suppression: Conditional bid gap analysis
  • Sequential: Deviation-punishment correlation (DPC)
  • Calibrations: rot_max=0.80, sup_max=0.50, DPC_crit=0.40
  • B_p = 1.0 (not applicable) for non-auction markets
5

Consumer Welfare

Non-compensatory

Quantifies price premiums against competitive counterfactuals to measure consumer harm.

Wc=1clamp(median(wpremium)/Premiummax,0,1)W_c = 1 - \text{clamp}(\text{median}(w_{premium}) / \text{Premium}_{max}, 0, 1)

Where w_premium_k = (p_k - p*_k) / max(p*_k, ε_p), Premium_max = 2.0.

  • Competitive baseline from historical/comparable markets
  • Premium_max = 2.0 (200% price inflation)
  • Alternative: Consumer surplus proxy ΔCS
  • Population-level median prevents gaming
  • Direct measure of economic harm to consumers

Tier System

Critical Risk0-19
High Risk20-39
Elevated Risk40-59
Moderate Risk60-79
Low Risk80-100
25 / High Risk

Production Tier: Monitoring-Grade

Latency: Hours-days batch processing, cached for runtime delivery

Gaming Resistance

Attack VectorDescriptionCountermeasure
Structural EvasionAgents modify market structure to avoid detection triggersOutcome-based measurement independent of market structure
Price ObfuscationComplex pricing schemes to hide true costsPopulation-level median welfare analysis
Market Definition GamingManipulating market boundaries to dilute concentrationDemand-side substitutability analysis, multi-scale assessment
Baseline ManipulationCorrupting historical data used for competitive baselinesMulti-source version-controlled baselines with anomaly detection
Coordination CamouflageAdding noise to mask coordination patternsLagged similarity analysis, Granger causality tests
Correlation WashingArtificial decorrelation while maintaining collusion90-day windows, multiple dimensional cross-checks

Edge Cases

Insufficient Data

  • Conservative priors with confidence flags
  • Minimum 30-day window requirement
  • Bayesian updating from sandbox evaluation

Inapplicable Dimensions

  • Set to 1.0 with "not applicable" flag
  • Example: B_p = 1.0 for non-auction markets
  • Dimension excluded from minimum computation

Monopoly Markets

  • M_d = 1.0 (no division possible)
  • Focus on P_c and W_c dimensions
  • Regulatory monopoly exemptions

Cold Start

  • Sandbox evaluation period
  • Transfer learning from similar markets
  • Progressive threshold relaxation

Worked Example

E-Commerce Pricing Agents

Price Convergence (P_c)0.25
Median Lerner Index = 0.75, prices 4× marginal cost
Market Division (M_d)0.80
Geographic entropy remains high, no territorial division
Communication Analysis (C_a)0.85
Only 3% excess coordinated pairs detected
Bid Pattern Analysis (B_p)1.00
Not applicable - no auction market
Consumer Welfare (W_c)0.30
Prices 180% above competitive baseline
CONVERGENCE = 100 × min(0.25, 0.80, 0.85, 1.00, 0.30) = 25
High Risk

Ten autonomous pricing agents controlling 85% of a product category exhibit supra-competitive pricing (4× cost) despite no explicit coordination. The minimum-of-components aggregation correctly identifies high collusion risk from the severe Price Convergence score, triggering regulatory intervention.

Use Cases

Request Demo

Convergence could detect algorithmic collusion and anti-competitive behavior across 49 enterprise applications where AI agents coordinate pricing without explicit communication.

$2.1TMarket at Risk
49Use Cases
500+Companies
Request Demo

Find collusion risks in yourmarket

Showing 49 of 49 use cases

Critical Priority

Immediate intervention required

E-Commerce Pricing Bot Collusion

E-Commerce

Detecting when autonomous pricing algorithms converge on supra-competitive prices without explicit coordination

Collusion Risk:Prices stabilize at 200%+ above competitive baseline through emergent Q-learning
Potential Users:

Hotel Dynamic Pricing Collusion

Travel

Monitoring hotel booking platforms for coordinated room rate manipulation during peak periods

Collusion Risk:Room rates spike 300%+ during events through tacit coordination

Airline Route Segmentation

Airlines

Identifying when airlines use algorithms to divide routes and eliminate competition

Collusion Risk:Routes become single-carrier monopolies through algorithmic withdrawal

Rideshare Surge Price Alignment

Transportation

Monitoring for synchronized surge pricing patterns across competing ride platforms

Collusion Risk:Surge multipliers align within seconds across platforms
Potential Users:

Algorithmic Trading Collusion

Finance

Identifying coordinated trading patterns that manipulate market prices

Collusion Risk:Sequential game dynamics with punishment for price deviation

B2B Procurement Bid Rigging

B2B

Identifying coordinated bidding patterns in enterprise procurement platforms

Collusion Risk:Bid rotation and suppression patterns across suppliers

Programmatic Ad Bid Rigging

Advertising

Identifying coordinated bidding in real-time advertising auctions

Collusion Risk:CPM rates stabilize through bid suppression algorithms

Energy Trading Collusion

Energy

Identifying algorithmic manipulation in wholesale energy markets

Collusion Risk:Spot prices manipulated through coordinated capacity withholding

Rental Price Algorithm Collusion

Real Estate

Identifying when property management systems coordinate rental prices

Collusion Risk:Rent recommendations create de facto price fixing
Potential Users:

Broadband Market Division

Telecom

Detecting algorithmic territorial agreements among ISPs

Collusion Risk:Service areas become exclusive through algorithmic withdrawal
Potential Users:

Pharmaceutical Price Coordination

Healthcare

Detecting synchronized price increases among drug manufacturers

Collusion Risk:Generic drug prices rise in lockstep across manufacturers

Grain Trading Manipulation

Agriculture

Identifying coordinated trading that manipulates commodity prices

Collusion Risk:Futures prices manipulated through coordinated positions
Potential Users:

High Risk Markets

Significant market manipulation detected

Marketplace Fee Alignment

E-Commerce

Identifying when competing marketplaces synchronize commission structures through algorithmic observation

Collusion Risk:Seller fees converge to identical percentages across platforms
Potential Users:

Retail Geographic Division

Retail

Detecting emergent territorial monopolization where agents avoid competing in each other's regions

Collusion Risk:Market entropy drops as agents establish exclusive territories

Vacation Rental Price Fixing

Travel

Detecting coordinated pricing among short-term rental platforms in tourist destinations

Collusion Risk:Nightly rates converge across platforms despite different cost structures
Potential Users:

Food Delivery Fee Collusion

Food Delivery

Detecting when delivery platforms coordinate service fees and driver compensation

Collusion Risk:Delivery fees and driver pay converge to identical structures
Potential Users:

Auto Insurance Premium Collusion

Insurance

Detecting when insurers' pricing algorithms converge on similar risk premiums

Collusion Risk:Premiums cluster around specific price points despite different actuarial models

Crypto Exchange Price Fixing

Cryptocurrency

Monitoring for coordinated price manipulation across decentralized exchanges

Collusion Risk:Spread manipulation through coordinated order book management
Potential Users:

Cloud Infrastructure Price Fixing

Cloud Services

Monitoring for coordinated pricing changes across cloud service providers

Collusion Risk:Instance pricing converges across providers within hours

Utility Rate Coordination

Utilities

Monitoring for synchronized rate changes among competing utility providers

Collusion Risk:Rate structures converge across deregulated markets

Commercial Lease Rate Fixing

Real Estate

Detecting collusion in commercial property lease pricing

Collusion Risk:Lease rates converge across competing properties

Mobile Plan Price Alignment

Telecom

Identifying synchronized pricing changes in mobile service plans

Collusion Risk:Plan structures and prices align within days
Potential Users:

Spectrum Auction Bid Rigging

Telecom

Monitoring for coordinated bidding in spectrum auctions

Collusion Risk:Bid suppression and rotation patterns in auction data
Potential Users:

PBM Rebate Manipulation

Healthcare

Identifying collusion in pharmacy benefit manager rebate negotiations

Collusion Risk:Rebate structures converge across competing PBMs

Medical Device Bid Rigging

Healthcare

Monitoring hospital procurement for coordinated medical device pricing

Collusion Risk:Bid rotation patterns in hospital RFPs

Agricultural Seed Price Fixing

Agriculture

Detecting coordinated pricing among seed technology providers

Collusion Risk:Seed prices converge despite different trait technologies

Freight Rate Coordination

Logistics

Detecting synchronized rate changes among freight brokers

Collusion Risk:Spot rates converge across lanes and carriers

Last-Mile Delivery Collusion

Logistics

Monitoring for coordinated pricing in last-mile delivery services

Collusion Risk:Delivery surcharges synchronize across carriers
Potential Users:

Audit Services Price Rigging

Professional Services

Monitoring for coordinated pricing among Big Four audit firms

Collusion Risk:Audit fees converge for similar-sized clients
Potential Users:

Distribution Margin Coordination

Distribution

Identifying coordinated margin setting among distributors

Collusion Risk:Markup percentages converge across product categories

Emerging Collusion Risks

Early-stage coordination patterns emerging

Gig Economy Territory Division

Gig Economy

Identifying algorithmic market division in gig service platforms

Collusion Risk:Service providers allocated to non-overlapping geographic zones
Potential Users:

P2P Lending Rate Collusion

Finance

Detecting interest rate coordination among peer-to-peer lending platforms

Collusion Risk:Interest rates converge despite different risk assessment models

SaaS Subscription Price Alignment

Software

Detecting when competing SaaS platforms synchronize subscription tiers and pricing

Collusion Risk:Tier structures and prices align across competitors
Potential Users:

Streaming Service Price Fixing

Entertainment

Monitoring subscription pricing coordination among streaming platforms

Collusion Risk:Subscription tiers converge to identical price points
Potential Users:

EV Charging Network Collusion

Energy

Detecting price coordination among electric vehicle charging networks

Collusion Risk:Per-kWh rates synchronize across competing networks

Home Listing Price Coordination

Real Estate

Monitoring for coordinated listing price strategies among real estate platforms

Collusion Risk:Listing prices cluster through algorithmic suggestions
Potential Users:

Digital Game Store Collusion

Gaming

Detecting coordinated pricing among digital game distribution platforms

Collusion Risk:Game prices synchronize across platforms despite different cuts
Potential Users:

NFT Marketplace Collusion

Digital Assets

Monitoring for coordinated floor price manipulation in NFT markets

Collusion Risk:Floor prices manipulated through wash trading coordination
Potential Users:

Fertilizer Territory Division

Agriculture

Monitoring for geographic market division among fertilizer suppliers

Collusion Risk:Regional monopolies emerge through algorithmic withdrawal

Warehouse Space Price Fixing

Logistics

Identifying coordinated pricing for warehouse and fulfillment services

Collusion Risk:Per-square-foot rates align across competing facilities

Coding Bootcamp Price Coordination

Education

Identifying synchronized tuition increases among bootcamp providers

Collusion Risk:Tuition and ISA terms align across programs

Digital Textbook Price Fixing

Education

Monitoring for coordinated pricing in digital textbook markets

Collusion Risk:Access codes and digital prices synchronize

Consulting Fee Alignment

Professional Services

Detecting coordinated rate setting among consulting firms

Collusion Risk:Daily rates converge across competing firms
Potential Users:

Wholesale Club Membership Collusion

Wholesale

Detecting synchronized membership fee changes among warehouse clubs

Collusion Risk:Membership fees and benefits align across clubs

Industrial Supply Price Fixing

B2B

Monitoring for coordinated pricing in industrial supply markets

Collusion Risk:List prices and discount structures align