A body of case law can be made to look an awful lot like a solar system, with clear centers of gravity, precipitating events, dependent relationships — and random moons just off doing their own things.
With predictive coding-related law, there was not so much a deliberate crawl from the ooze as a sudden big bang: In February 2012, the seminal Da Silva Moore case condensed all the formative matter that had existed in the legal, academic and popular ethers into the first tangible judicial ruling validating the then-fledgling method; an explosion of legal rulings surged forth in its wake. No fewer than two dozen substantive decisions dealing with predictive coding — generally approving its use or signing off on specific protocols — have arisen from federal, state and tax courts over the three and a half years since Da Silva Moore.
This evolving universe, no doubt, has its quirks. Courts have cited, among other sources, vendor-authored white papers, student-prepared academic studies, and newspaper articles as authorities substantiating predictive coding’s superiority or parity vis a vis other search methods — or underscoring flaws with those alternative methods.
Below is a first attempt — a working draft — at a visual representation of the evolution of predictive coding case law in the United States. The purpose of this exercise is to illuminate the rationale and authorities underpinning judicial validation of that method (though some say it does not need validation by the courts in the first place), and to show the interplay of law, academia and opinion in the formation of precedent.
- The graphic is linear in time, moving from left to right.
- Dark green circles denote judicial rulings dealing with predictive coding.
- Light green circles denote judicial rulings that do not — and in this instance, deal specifically with the weaknesses of keyword searches.
- Blue circles denote academic studies or articles.
- Orange circles denote non-academic articles, or any writing that was not published in an academic, peer-reviewed publication.
Citations are shown by connecting lines. Lines should be traced to the left to show authorities (rulings, articles, etc.) to which a case cites. For example, Da Silva Moore cites to the Blair and Maron study, a Legal Tech News article called “Search, Forward,” and the O’Keefe decision, among other authorities.
The more lines a circle has connecting to it on its right side, the more it can be said to be influential. For example, 10 rulings have cited Da Silva Moore, as shown by the 10 lines originating from that circle’s right side.
The size of the circle corresponds to the number of times other cases (or foundational articles) cite to it, and, thus, its relative influence. Bigger circles can be said to be more influential in terms of number of subsequents cases that cite to them.
This post reviews all known substantive rulings dealing with predictive coding — or at least those known to the author — as of September 22, 2015. Readers are encouraged to vet this work and to submit other cases that are not discussed below.
Note that this analysis does not account for rulings that have since been overturned by an appellate court, though none are currently known to the author. Such opinions, should they exist, will be covered in subsequent posts. A compendium of cases, which are also available at Ctrlinitiative.com, are available below in a downloadable spreadsheet. The spreadsheet charts citations and also includes links to all the rulings, articles and studies.
To be sure, there is a better visual vehicle to represent case law and readers are encouraged to identify it.
Again, please note that this document is a working draft. If you notice any errors or know of other substantive rulings that are absent, please contact Robert Hilson at firstname.lastname@example.org.