Potentials of computational antitrust 

1. Investigation of anti-competitive practices 
2. Merger control 
3. Design and monitoring of anti-trust policies 

Anti-competitive practices (1) 

• Antitrust agencies mostly rely on reactive methods to detect collusion 
  • e.g. leniency applications; screening tools 
• Technology (e.g. AI, blockchain) can help market players implement and sustain anticompetitive practices 
• \(\to\) we want to use computational tools that provide us proactive response 
  • Machine learning will be helpful 

ML Example 

In Machine learning with screens for detecting bid-rigging cartels, by Martin Huber and David Imhof: 

• Combined ML techniques with statistical screens for predicting collusions in tender procedures within the Swiss construction sector 
• This approach can be integrated in a process of ex-ante detection of collusion 

Anti-competitive practices (2) 

• Computational tools allow law enforcers to analyze documents faster 
  • E.g. European commission used computational tools to study the 1.7 billion search queries for its investigation in the Google Shopping case. 
• Computational tools enables market players to conduct more thorough internal audits 
  • given a tool for assessing compliance with antitrust laws 

Merger control (1) 

• Agencies must make choices under time constraint \(\to\) greater uncertainty 
• Computational tools can help analyze extensive data sets within the time constraint 

Merger control (2) 

• Information asymmetry between companies and agencies 
  • Companies may choose to not disclose or disclose incorrect information 
  • Agencies do not send back processed information to companies 
• Computational tools can introduce a communication tool that ensures data integrity 

Antitrust policies 

• Computational tools can provide better quantitative retrospective analysis of antitrust investigations, merger control decisions, and public policies. 
  • research: more empirical studies on digital markets 
• Through simulations, computational tools can provide us predictive insights about the effects of new public policies and legislation. 

ML example 

From Gleaning Insight from Antitrust Cases using Machine Learning by Giovanna Massarotto and Ashwin Ittoo 

• unsupervised model 
  • clusters cases based on similarity
• Algorithm discovers underlying patterns between 72 cases from 2005 to 2019
• cases from the data and computer industry were generally clustered with those in the healthcare industry \(\to\) that these industries raise similar antitrust concerns

Potentials of computation tools 

1. Investigation of anti-competitive practices 
   1. reactive \(\to\) proactive; more efficient processing; more thorough internal audits 
2. Merger control
   1. better communication and better tools for processing data under time constraint  
3. Designing and monitoring of anti-trust policies 
   1. better retrospective analysis and predictive insights 

Challenges of computational antitrust

1. General challenges of computational law 
2. Specific challenges to computational antitrust 
3. The human challenge 

General challenges 

• Who's in control? 🤖 or 👨👩 ?
• Transparency is key
  • One can reveal the process of computational decision-making. 
  • One can design a computational tool that emphasizes transparency 
• For the broad adoption of computational tools, transparency seems to be a prerequisite. 

Specific Challenges (1) 

• Developing the right tools 
  • Tools that assess compliance with antitrust laws and rulings 
  • Tools that automate enforcement and merger control procedures 
  • Testing of such tools

Specific Challenges (2) 

• Data and its scope - what kind of data do we need for the tools we created? 
• \(\to\) What kind of tools are we creating? 
  • retrospective analysis ✅ 
  • counterfactual analysis ✅
  • predictive analysis❓ ​ 
    • can we justify ex-ante regulations? 

Specific Challenges (3) 

• The role of computational tools in decision-making processes: 
  • To what extent can they justify decisions on their own? \(\to\) need to hire personnel 
  • What is the (comparative) importance of non-computable elements? 

A Human Question 

• Most critical challenge: interaction between legal system and technical tools \(\to\) this is a human challenge 
• \(\to\) We need more cooperation between fields and institutional changes 
  • provide proper incentives 
  • establish conditions for sustained collaboration 

Challenges of computational antitrust 

• General challenges to computational law 
  • \(\to\) Ensure transparency 
• Specific challenges to computational law 
  • Developing and testing of tools 
  • Obtaining data 
  • Accounting for non-computable elements 
• The human challenge  
  • Encourage cooperation and institutional changes