This article analyses the outcomes of huge language mannequin processing to disclose a number of essential patterns within the feedback to the draft merger tips.
In December 2023, the U.S. Department of Justice (DOJ) and the Federal Trade Commission (FTC) launched new merger tips, considerably updating their acknowledged strategy to mergers. The launch of those tips concluded a course of that started in the summertime of 2022 with the discharge of a request for info (RFI) that requested for enter on a variety of questions and priorities associated to merger enforcement, and continued via the discharge of draft merger tips (DMGs) in July 2023.
A complete of 1,906 feedback had been submitted after the discharge of the 2023 DMGs, of which 1,689 might be categorised as coming from members of most people. These feedback didn’t supply technical strategies for the rules however as an alternative pointed to common areas of concern, reminiscent of larger costs, or particular industries, reminiscent of telecoms.
Modern machine studying methods, together with giant language fashions (LLMs), permit us to look at these feedback to find out whether or not commenters favor better enforcement and what industries or matters they imagine want better scrutiny. Cornerstone Research’s Data Science Center professionals quantized this mannequin to run effectively from fully inside our safe knowledge heart. A evaluation utilizing this LLM method confirmed that greater than 90% of feedback supported better enforcement, whereas solely 2% might be categorised as supporting much less enforcement.
In this text, Andrew Sfekas, with assist from the Data Science Center, utilized this LLM method to disclose a number of essential patterns in how and the place commenters wished better merger enforcement.
The article was initially revealed in Law360 in February 2024.