The arrival of AI may be about the revolutionise our lives. Claims are made that programmes such as ChatGPT and Bard will fundamentally change our world as much as the industrial revolution and the IT revolution did. Sceptics claim that AI may be about to steal people’s jobs, or even that it might threaten the continued existence of humankind.
Sir Geoffrey Vos, the Master of the Rolls (the head of the civil justice system) has suggested there is “a real possibility that AI may become more intelligent and capable than humans” and that before too long there will be AI judges. Initially AI might only handle minor straightforward issues, but with growing public confidence, it may one day be used to make much significant decisions.
For someone who makes his living from working in the law, this is a discomforting prediction. No doubt, if AI judges can make decisions, there will also be AI lawyers stealing my business.
Will clients be willing to put their legal matters in the hands of AI? Will AI be able to cope with the complex application of the law? Knowing the law is not enough to be a good lawyer. If it was, law graduates fresh out of university or law school would be able to practise as solicitors immediately, rather than undergo a training contract. Of course, AI will be able to learn, but learning how to apply the law is a subtle skill and involves much exercising of personal judgement.
There is currently no formula or algorithm to decide how to divide a couple’s income and assets in a divorce, (apart from assessment of child maintenance). The government has recently asked the Law Commission to consider changing the law to make outcomes easier to predict, which would probably tie the judge’s hands and would, I suggest, lead to outcomes that are easier to predict, but less likely to actually meet the needs of the parties. An AI judge would have to reply on a formula or algorithm.
I often tell clients that dividing income and assets in a divorce is not a precise science, it is more of an art. (The family solicitor who trained me back in the mid 1990’s once memorably described it to me as being a dark art). I recall that learning how to predict what a court might order in a financial case and advising clients on what they see should seek or offer was incredibly hard. No two cases are the same. The Matrimonial Causes Act 1973 provides a long list of factors that the court should take into account, but it is really quite vague. Reported court decisions indicate how these the factors should be applied, but the subject is vast. Most cases probably do not have a single possible outcome; there are a range of possible outcomes which are within the range of what a court might order. If you placed the same set of facts before two judges, you will probably get two different outcomes. They ought to be similar – but they are unlikely to be identical.
I can see AI being used to deal with undisputed divorce applications. Divorce (by which I mean the process of ending a marriage, not the financial disputes that spin off from it) could be handled by AI. No-fault divorce is straightforward and increasingly more of an administrative process rather than a judicial one. Former President of the Family Division, Sir James Munby suggested that divorce should be handled by a Registrar of Births, Deaths, Marriages and Divorces, not by the Family Court. Although district judges still formally pronounce the Conditional Orders, the actual granting of the certificate of entitlement to a divorce beforehand is now handled by the court’s Legal Advisers (what used to be called magistrates clerks). They are lawyers whose role was traditionally to advise lay magistrates, but who often now perform very low-level judicial functions as well in the Family Court. It is easy to see them being replaced by AI. Divorce applications should be successful; if the applications are made properly, AI could easily check the applications and ensure that everything is in order.
AI could perhaps also deal with applications for financial consent orders. Where a couple reaches an agreement about their finances, they need to seek a financial consent order from the Court that reflects the terms of their agreement. Solicitors draft the order and submit it to the court with a financial statement so that the judge can considered whether the terms of the agreement and the order are appropriate. Could AI approve or reject orders? If the paperwork is completed properly with a clear justification of the terms is provided by the solicitors, an AI judge might find it easy to decide. I am aware that some research was done into how long judges take to consider these applications, and I was shocked to hear that most applications were approved in a matter of minutes. If human judges are spending that little time on it, would an AI judge be able to do it more accurately?
Could AI draft the order and the statement too? We are taking about quite a complex legal documents, albeit one that increasingly uses standard wording. Drafting these documents is a skilled job, but could AI draft it more quickly and less expensively than a solicitor? Perhaps.
If AI could learn to do that, then how long could it be before AI could learn to make a decision in a contested case, or even represent a client? It reminds me of a very old episode of Star Trek where Spock is on trial. The court has a computer that regular contradicts statements by witnesses, bot-splaining to everyone that their evidence is false.
A lot may depend on how quickly people would be prepared to come to terms with the idea of entrusting this task to AI. Look at self-driving cars. I think it’s fair to say that people don’t entirely trust them yet. The UK government has just granted Ford permission to activate its self-driving software on its electric Mustangs. The driver has to pay attention to the road or else the car grinds to a halt. It can only be use on motorways. People worry that that is not ready to go it alone yet. However, it is worth noting that self-driving cars have travelled vast distances during tests and have killed hardly anyone, a much better record than human beings. In fact, when a self-driving Uber test car killed a pedestrian, the car was exonerated whereas the human back up driver faced charges. Will AI become more reliable and more consistent than humans in handing down judgments? Will people be prepared to entrust decisions about their children to a piece of software?
Deciding arrangements for children is a very tricky decision for any judge to make. It can turn on the smallest of things. I recall a case where two parents were arguing about with which parent their young child should live. Both parents had their flaws but appeared equally capable of looking after their child. There was nothing obvious that might tip the scale in one party’s favour. I suspect that an AI judge might have made what was then known as a shared residence order so that the child spent half of the time with each parent. Give their inability to co-operate on any level, such an outcome would have been a disaster. In the end, the judge noted that one parent was illiterate, but the other could read. As the child was about to start school and would be learning how to read, the judge decided that the parent who could read would be better placed to be the child’s primary carer. Would an AI judge have been able to make that decision? I suspect not.
We shall have to wait and see what the future brings. For those of us worried about our livelihoods, there is solace to be had in a recent encounter between the Chancellor of the Exchequer and an AI bot. Jeremy Hunt asked the bot how he thought that he was managing as Chancellor so far. The bot told him “Jeremy Hunt is not the Chancellor of the Exchequer, that’s Rishi Sunak.” Clearly the bot had not been keeping up with past year’s turbulent politics.
The future is difficult to predict. The current excitable stuff in the media about AI may be massively exaggerating the possible impact. Nevertheless, dare we be complacent about this? About 60 years ago, the head of IBM famously predicted that the world would only sustain a handful of computers. There are now probably more computers on the planet than humans. There are definitely more in the developed world. As I sit writing this blog in a coffeeshop, I have four computers with me; my laptop, my iPad, my iPhone and my smart watch.
Clearly AI is coming and is likely to be used in the law. The question is for what? Simple stuff or complex stuff? How will lawyers use it? Could AI have drafted this blog for me? In fact, how do you know that it didn’t?
22 April 2023
If you would like to arrange a consultation, please telephone 01206 848426 or click here.