Originally published in Persian on 23 February 2023. This essay presents the author’s views at that time; artificial intelligence has continued to develop since it was written.
Will artificial intelligence replace lawyers? Could a judicial system be built around it? These questions are being widely discussed. I would like to see that day, but I believe it is still a long way off. I hope to write more about the subject.
Should lawyers fear AI? Should we compete with it? Where do human beings have an advantage over machines, and where do machines have an advantage over us?
ChatGPT changed our world. Even people already working in machine learning found it remarkable; for ordinary users, beginning a conversation with it was a striking moment.
I do not intend to explain how machine learning works. But it is important to consider its limitations and compare them with human thinking. Here I want to explore how AI might affect legal work.
Where are we now?
First, we have clearly lost the contest in calculation. Computers can perform thousands of calculations in a second, far beyond our ability. Work that consists chiefly of such calculations can therefore be automated. Suppose a man dies leaving a wife, two sons, a house, a piece of land and a bank loan of 100 million tomans. How should the inheritance be divided? A small program could perform the relevant calculation. Wherever legal work comes down to computation, computers can help.
Second, machines can learn from examples. Show a system many images labelled “cat” and it can learn to identify cats. Likewise, exposure to many contracts can help a system produce a draft. Does that mean AI will replace contract drafting?
Negotiating and drafting contracts
A contract has several stages. First, the parties negotiate their transaction. In my view at the time of writing, AI could not conduct that human negotiation for them. So if you are a contract negotiator, I do not think you need to fear for your work.
Next comes drafting. AI can generate a contract based on instructions. But standard forms already existed: organisations such as FIDIC publish them, yet contract drafters continue to work. AI may make drafting easier, but it is difficult to say that it can be done without human supervision. Even when one contract resembles another, each transaction has differences, however small. Drafting teams may become smaller, but I do not expect them to disappear.
Research and legal reasoning
Before bringing a claim or beginning a dispute-resolution process, lawyers research relevant decisions and legislation to build an argument. AI may help find cases in common-law systems and provisions in civil-law codes, as well as arguments supporting a position. My view in 2023 was that applying those materials to the particular case remained a human task.
One familiar way to structure legal writing is IRAC: issue, rule, application and conclusion. I saw AI as useful for finding rules, while identifying the issue, applying the rule and reaching a conclusion remained with us. These are judgments about the technology at the time, not a substitute for checking any research against authoritative sources.
Litigation and contested facts
If a car hits a tree, there may be three accounts: those of the driver, the tree and an observer. A confession might appear to make a case straightforward, but real cases are often ambiguous. One suspect may say that he killed another person in self-defence; the prosecutor may say the killing was planned. Which account is true?
A judge must decide whether to accept a witness’s evidence, and a witness may be challenged in court. When arguments conflict or the facts are uncertain, I question whether AI can take over the human task of evaluating them.
Mediation and trust
People differ from one another, and even the same person’s behaviour can change within an hour. A mediator must respond to different people and circumstances. The parties must also trust the mediator. In 2023, I doubted that many people would readily trust a robot in that role. Perhaps a later generation will be more comfortable communicating with machines, but for now many would hesitate to let AI take the lead.
Knowing that a mediator is a robot could itself undermine confidence. If that happens, the mediator loses the very tool on which the process depends: trust.
There is also a philosophical question. We still believe in the central place of human beings in the world. Accepting AI as a judge could mean accepting the primacy of the machine over the person. I do not think humanity is ready for that yet.
