This could get ugly. I’ll step our way through it so stay close and hopefully you will make it through to the end okay.
Dr. Stephen Wolfram is the guy you did not hang around with when you were in school. He was born in London in 1959. As often happens with people of high intelligence, he struggled in school and had no patience for the “silly” arithmetic books he was asked to read. But by his early teens, he had written three books on particle physics (not published).
One of the reasons you would not have hung around with Stephen in school is that he hardly spent enough time at a school for you to get to know him. By age 15, he had published articles about his research in quantum field theory and particle physics. He went to Eton College, but left before graduating and at age 17 entered St. John’s College, Oxford. He also left St. John’s before graduating and enrolled at the California Institute of Technology where, at age 20, he received a PhD in particle physics. One of the members of his thesis committee was Richard Feynman (yes, that Richard Feynman). What next? He joined the faculty at Caltech and at age 21, became the youngest recipient of a MacArthur Fellowship (the so-called genius grant).
If you think Richard Feynman was a brilliant theoretical physicist who did things ranging from assisting in the development of the atomic bomb, to creating Feynman diagrams (visual representations of the mathematical expressions describing the behavior of subatomic particles), to nanotechnology, you are right. But he also was perceptive about human character. When Wolfram wrote to Feynman saying he was considering starting an institute to study complex systems, Feynman replied “You do not understand ordinary people,” and suggested Wolfram “find a way to do your research with as little contact with non-technical people as possible.” Again, another reason why you probably would not have hung with Wolfram.
Wolfram left Caltech and joined the faculty of the University of Illinois at Urbana-Champaign where he founded the Center for Complex Systems Research and the journal Complex Systems. When he was at Caltech, Wolfram had developed a computer program called Symbolic Manipulation Program. A battle with Caltech over the rights to the program and related issues led to Wolfram leaving for the University of Illinois. Shortly after arriving at Illinois, Wolfram began developing Mathematica and within a year founded his company Wolfram Research. Today, Mathematica is widely used around the world and Wolfram Research, which Wolfram joined full-time shortly after founding it, develops and promotes the program.
In 2002, Wolfram published the book A New Kind of Science, in which he argues that the universe is digital. He further argues that simple computational systems can be devised to model and explain all of nature. In 2014, Wolfram finally named the programming langue that had been driving Mathematica for 25 years, calling it “Wolfram Language.” Wolfram Language can be used to write the computational systems, but Wolfram had been expanding the Language’s reach. Wolfram spends his time on Mathematica, on developing Wolfram Language, and on giving it greater exposure so others will use it. In essence, Wolfram followed Richard Feynman’s advice by creating a world in which he can spend most of his time working with technical people on his vision of a computational future.
And I Care Because …?
Last week, Wolfram posted a long blog post laying out his vision for computational law. The post covers a lot of ground and stretches from Aristotle to the present, so I won’t try to cover it all in my recapitulation. Instead, I’ll focus the rest of this blog on the key point in Wolfram’s blog, his argument that now is the time (and of course Wolfram Language is the vehicle) for creating a symbolic discourse language. In other words, Wolfram believes we are ready for a language we can use to express legal concepts and which computers can use to compute outputs. Creating the symbolic discourse language, within Wolfram Language (a symbolic language) is his next step. Again, talk like this is probably another reason why you wouldn’t have hung with young Wolfram.
Think of symbolic discourse language as something that exists between natural language and computer language. Without getting deeply into computer software and hardware, think of the computer’s operating system (e.g. Windows or Mac OS) as the base level. On top of the operating system we have applications, like Word. When you want to write a letter, you can open Word and just type. Word interacts with the computer’s operating system and the operating system interacts with the hardware, so that when you click “print” your letter is printed.
That system worked well for lawyers and poets, but those who used math were left struggling. They had to program the computer to run their computations, and that meant learning computer languages such as Fortran (in the old days) or C.
Wolfram created a new language that allowed people to run math and get answers to formulas or graphs without having to go deep into programming. The new language, Wolfram Language, is a symbolic language. That means you can enter relatively simple commands and Wolfram Language converts them into the complex commands that drive the computer. The more sophisticated the language, the more symbolic the commands you can use.
If you ask Wolfram Alpha, which takes as one form of input natural language, “what is the diameter of the earth?” it can translate your natural language inquiry into the code needed to search for the information, assemble it, and present it to you in a way that you can understand.
Now think of a court decision. Judges do not use symbolic language. They attempt to explain the law, the facts, and their reasoning using natural language. But using natural language can get messy. Think about separating “preponderance of the evidence” from “beyond reasonable doubt.” You get the terms, but that doesn’t mean a computer or others get the terms. They convey a concept, but not precisely.
A symbolic language could take each term and turn it into something a computer can understand (e.g. >50.0%). Once the computer can understand it, it can receive inputs and deliver outputs. Lawyers and judges would then write contracts, briefs, case law, and other materials using the symbolic discourse language instead of natural language.
If you are straining to extend this idea to all legal discourse, that isn’t surprising. It will take quite an effort to develop the entire symbolic discourse language. But Wolfram’s point is that our knowledge and tools have developed to the point where he thinks his team can do it.
Don’t Get Rid Of Lawyers Just Yet
Let’s address the first issues that come up in a lawyer’s mind when reading this story: what is good or bad in it for me? You may find it surprising, but Wolfram does not take the position that the symbolic discourse language will be the end of lawyers. He says, “Today lawyers have to learn to write legalese. In the future, they’re going to have to learn to write what amounts to code: contracts expressed precisely in a symbolic discourse language. … [I]t will help lawyers think better about contracts.” For those in legal education, this is another, and perhaps the most powerful yet, reason to start teaching law students logic and coding.
If symbolic discourse language won’t decimate lawyers, will it decimate the law. Will law become so simple that anyone can do it? Not so, according to Wolfram,
Once computational law becomes established, the complexity of what can be done will increase rapidly. Typically a contract defines some model of the world, and specifies what should happen in different situations. Today the logical and algorithmic structure of models defined by contracts still tends to be fairly simple. But with computational contracts it’ll be feasible for them to be much more complex—so that they can for example more faithfully capture how the world works.
He goes on to describe how the symbolic discourse language will interact with machine learning software that is gathering information from other sources (e.g., the internet) that the language uses to inform the contract. This gets a bit tricky, but I’ll take a stab at explaining it borrowing from one of Wolfram’s examples.
The contract calls for X to happen when condition Y is satisfied. But Y is something itself difficult to define as “satisfied” or “not satisfied” in simple terms. Wolfram uses the example of fruit. I will pay you $10,000 for delivering to me a certain quantity of fruit meeting the standard “Fancy Grade.” The question is whether the fruit met the standard.
We could define the standard as no more than Z% of the fruit has blemishes and we can further define a “blemish”. A computer could examine all the fruit, calculate the percentage of blemished area, and feed that into the contract yielding an output: pay or don’t pay.
Many lawyers may be shouting “huzzah” right now. We’ve just said that law will evolve to a symbolic discourse language (in other words, legalese of a different type), become more complex, and require knowledge of both legal principles and computers. Is law going back to an opaque art that will require clients to pay for access? I don’t think so, but let’s leave that question to the side and explore other “what does it mean” questions.
Crushing Poetry Out Of Law
Every law student knows the Aristotle quote, “The law is reason, free from passion.” Wolfram says that symbolic discourse language would take us there, “In a sense, the symbolic discourse language is a representation in which all the nuance and ‘poetry’ have been ‘crushed’ out of the natural language.” This will raise some interesting questions, particularly when it comes to equitable considerations. Should contract law be devoid of poetry?
Going in another direction, we can ask how symbolic discourse language might affect our understanding of the economic underpinnings of contracts. On October 10, 2016, the Royal Swedish Academy of Sciences awarded the Sveriges Riskbank Prize in Economic Sciences in Memory of Alfred Nobel 2016 to Oliver Hart and Bengt Holmström “for their contributions to contract theory.” They have focused much of their work on the area of incomplete contracts. The theory starts with the thesis that contracts are incomplete, because they cannot specify what is to be done in every future situation.
Part of specifying the future is data, part is computational power, and part is complexity of the contract. Today, we can’t possibly analyze sufficient data to predict all possible future consequences. Even if we could get enough data, we would need tremendous computational power to analyze it. Finally, writing a contract to cover all the contingencies would result in a document no one would dare write or read.
Wolfram posits a future where those problems would be greatly mitigated. Computers can scour vast databases and use machine learning to analyze data relevant to the contract. With access to tremendous computing capacity, the power to analyze the data becomes available. Finally, if the contract will be written in code—and given that the computer itself could write at least some of that code—we don’t care how long the contract becomes. We can see a touch of this today in electronic trading on the stock exchange. Computers gather and analyze the data, develop algorithms based on the data, and place the trades.
As you can begin to see through the fog, data plays an ever more important role in contracting. Data helps inform the terms of the contract, but data also becomes the fodder for the programs that determine whether terms of the contract have been met. Data also affects the dispute resolution process. If both parties and the court have access to massive quantities of data and the computing power and systems (machine learning) to analyze the data, dispute resolution could become focused on very narrow issues as more contract issues are answered through complex contracts and data.
We Still Have Ground To Cover
Wolfram’s post (which consumes 20 single-space pages) touches on some of these issues and addresses many more, yet still leaves large gaps in its wake. The core proposition is this. Work so far, with some exceptions (many of which Wolfram notes) has been focused on backing into discourse analysis by examining what courts have done and attempting to find ex ante ways to construct systems for describing the logic of law. Wolfram proposes to construct a symbolic discourse language that lawyers, judges, legislators, and society would use to create law. Computers could use the language, augmented by machine learning analysis of relevant data, to evaluate questions arising under the law.
Wolfram acknowledges the huge amount of work it will take to accomplish this feat. But, as his biography suggests, he is not someone to shy away from challenging questions or large amounts of work.
Lawyers should consider what Wolfram proposes in a different light. Perhaps Wolfram will succeed or perhaps this will be a challenge that survives him. But, most of the work we see today involving computers and law involves attempts to automate the present or to decipher the past, not create the future. Just as we are at an inflection point in the delivery of legal services, one could argue we are at a similar point in the substance of law. It has become too chaotic and faces challenges too great (e.g., explosion of relevant data, speed of societal change), for the current approaches to developing law to work. Add on to that other issues, such as the privatization of law and many issues never reach the courts, affecting the evolution of common law.
Wolfram’s path is not artificial intelligence in the law. It doesn’t remove lawyers from the equation (though that is theoretically still possible at some point). Instead, and at a closer point in time, it leverages the power of computers to use data, handle complexity, and make law (in theory) more precise (though at a cost to humanity, a topic for another post). Those are benefits we can deliver to clients and, ultimately, what makes his path intriguing. It will be up to lawyers to determine whether they let this turn out to be an ugly or beautiful path.