The New Forgery — Legal Works
Could you employ Oliver Wendell Holmes, Jr. to write your legal briefs?
Rembrandt Harmenszoon van Rijn was one of the master artists during the 1600s. His paintings include Cottages before a Stormy Sky (1641), The Night Watch (1642), and The Jewish Bride (c. 1666). His fame carries through to today. The museum in Amsterdam featuring his life and work, Rembrandt House Museum, is a popular tourist stop.
Rembrandt had many pupils which, at times, has caused a bit of confusion. Experts have debated whether Rembrandt or one of his pupils painted certain works. Rembrandt also attracted the darker side of the art world. Experts have reclassified some of the 600 works attributed to Rembrandt as fakes. They also declared some fakes authentic. If there is one thing on which we can all agree, it is that Rembrandt scholars disagree. It seems that, when it comes to art, declaring authentic from fake also is an art.
Literary forgeries exist, but present some different challenges when we move from signatures and letters to book-length works. The most recent and famous book forgery was The Hitler Diaries in the early 1980s. The newspaper Stern had purchased 60 small diaries from journalist Gerd Heidemann. The diaries supposedly were written by Hitler during his rise to power in Germany. The journalist claimed he had recovered them from a plane crash. After some excerpts were released and questions raised, a study by the German Archives proved that the diaries were badly done forgeries and the journalist and his accomplice went to jail.
Literary forgeries present some of the came challenges as paintings. Does the media match the time when the work supposedly originated? Does the backstory match known facts? Like paintings, literary forgeries also hinge on technique. The painting must have the brushstrokes and style of the painter’s known works, and books must have the handwriting and style of the author’s known works. In both cases the forger must have intimate knowledge of the original artist’s work and techniques if he hopes to pull off the forgery.
The computer’s ability to capture a master’s brushstrokes or an auteur’s style may be at the point where we are looking at the beginning of a new era in creativity and forgery. Digital, physical, human, and law blend in a way that creates the 21st century crook.
Criminals And Technology
Criminals have used technology for, well, as long as there have been criminals and technology. New technologies may give rise to interesting uses that open new opportunities for those on the other side of the law, but that alone is not terribly creative. What is new is that the computer has gotten good enough to mimic creative activities to the point where, in some instances, we can’t tell the difference between the original and the new.
Start with this quiz run by The New York Times. It tests your ability to classify written works as written by a human or a computer. Were you able to correctly identify which pieces were written by a computer and which by a human? Probably not. In fact, much of what you read if you are a sports fan or an corporate earnings aficionado probably was not written by a person. Narrative Science, the company started by journalist-academics from Northwestern University’s Medill School of Journalism, uses its algorithm-driven computers to turn out many of the core stories in sports and finance that you read on the web.
Paintings involve a few more hoops for a forger to jump through. The forger must match the paint, the brushstrokes, and the overall style of the artist. But through the combination of digital and physical, a group of techno-artists at advertising company J. Walter Thompson has created a new painting in the Rembrandt-style. Using a 3D printer to layer the paint and an algorithm-driven computer that had studied Rembrandt’s known works, the team was able to create a new painting in the Rembrandt-style. Not quite a forgery, but a significant step forward.
Lawyers who practice in the arcane field of art forgeries may find these stories mildly amusing. They may get more interested when they and other lawyers realize that the algorithms brought to bear on the art world could be focused on the legal world. How far are we from computers writing legal briefs? Not as far as you might think.
To begin, the algorithms in both the painting and writing stories were not starting from scratch. The computer-turned-painter was trained for 18 months on Rembrandt’s work. The computers who write sports and finance stories have trained reading stories by human writers. Both had good teachers.
Let’s put this in a specific case. Could a computer review the trial record, look at the exhibits, read the case law, and write a decision? No, computers are not close to taking over that task. But if we scale back the scope, the question becomes more interesting. This is how one of Narrative Science’s competitors, Heliograf, helps write stories for The Washington Post:
Editors create narrative templates for the stories, including key phrases that account for a variety of potential outcomes (from “Republicans retained control of the House” to “Democrats regained control of the House”), and then they hook Heliograf up to any source of structured data — in the case of the election, the data clearinghouse VoteSmart.org. The Heliograf software identifies the relevant data, matches it with the corresponding phrases in the template, merges them, and then publishes different versions across different platforms. The system can also alert reporters via Slack of any anomalies it finds in the data — for instance, wider margins than predicted — so they can investigate.
The Computer That Writes A Brief
At the risk of fueling the AI hype, let’s take an excursion using a thought experiment. As far as I know, the service I am going to describe does not exist. But ask yourself this question, “how long is the walk from what Heliograf does to writing briefs?”
We feed a computer decisions written by an appellate judge we particularly admire. Let’s say that judge is Oliver Wendell Holmes, Jr. We use those decisions, and the many letters and other documents we have from Justice Holmes’ pen, to teach the computer his writing style.
We do this today with law students. They read decisions and learn to parse the good from the bad. Computers, given enough decisions and some guidance, can learn the same. Whether the computer would be better or worse than a human, we don’t know (we do not have studies yet, or at least we do not have studies generally available). It would be an interesting test to compare decisions by the real Oliver Wendell Holmes, Jr. to ones written by our computer using the case record.
The next task would be training a computer to understand the relationships between the appellate briefs and the court’s decision. Some judges borrow heavily from the briefs filed in a case when writing their opinions, while others borrow sentences here and there. Computers can quickly determine where overlaps occur. A lawyer could do what an editor does at The Washington Post, providing information from the record that the computer would use. It would not be hard to supply the procedural history. The lawyer could also point the computer to the relevant facts.
But is machine learning sufficiently advanced that a computer, given the appropriate briefs, could write an opinion? Writing opinions clearly requires more than taking the briefs, picking out the relevant parts, and knitting the results together. Judges often rely on cases and other materials not cited by the parties. They use what is called “tacit knowledge.” This is the body of knowledge they have picked up from a wide range of sources and which they bring into play when deciding cases. Should the computer also be allowed to search the Internet for relevant “context” material?
Lawyers dismiss the idea of computers writing anything as science fiction. They accept automating. Using a computer to fill in blanks in a human-authored template is not writing, they say. Computers can complete templates and computers can ask questions and record answers used to complete the templates (again, not writing).
What most lawyers do not realize is that various parties are working on teaching computers to write, just as I have described above. We may not be there yet, but the journey has started. There are many technical hurdles between where we are and success, but no law of nature blocks computers from learning to write. Considering that many disputes do not require creative leaps in legal knowledge and often are won or lost on mundane facts and law, it seems there is an entry point for computers that can learn this skill.
Collaborate To Build The Future
We are not on the verge of computers becoming lawyers and we still have some ground to cover before computers can turn out even the simplest briefs or decisions. But, we are well into a world where computers can automate part of what lawyers do. Lawyers who believe otherwise face a dark future. There are many ways to avoid this future. One is through collaboration.
Lawyers, law departments, and even law firms should start partnering with colleagues in academia to focus resources on how to put computers to work partnering with lawyers. This is a fertile, and mostly ignored, area that provides many benefits. Lawyers will become more knowledgeable about and comfortable with the ways computers already can help their practices. Scholars will spend more time working on real problems and driving practical solutions. Students will have more chances to integrate what they do in school with the real world. Clients will benefit
This essay was written by a human, but on a computer.
If you enjoyed reading this article, please recommend (click the heart ) and share it to help others find it!
About: Ken is a speaker and author on innovation, leadership, and on the future of people, process, and technology. On Medium, he is a “Top 50” author on innovation and leadership. You can follow him on Twitter, connect with him on LinkedIn, and follow him on Facebook.