Tomorrow's Lawyer: Will Artificial Intelligence replace the human lawyer?

Maybe. But I don't see it catching on very quickly.

For straightforward procedural things like conveyancing, I can see it - although things like wills and trusts where there are long term consequences, I could see AI making a lot of mistakes that would cost people down the line. At some point, AI would have to stop being supervised by a human otherwise its pointless.

For anything contentious or criminal, the issue is that eventually the endpoint is not only AI solicitors and advocates, but AI judges - and no one will want that. Simply put, I don't see how an AI will ever be able to judge merit or harm. If you plead guilty to shoplifting, but say that in mitigation you had to feed your starving children - how will an AI decide if that is mitigation or just a transparent pack of lies? Humans always seem to err on the side of caution and have the tendency to give the maximum forgiveness to any given defendant or applicant. I don't see an AI doing that - you would get people getting judgements or sentences as per statute or desert - and that would make people very very unhappy. I'd give it 1 year before you would have demands for AI judges to be scrapped because AI is racist or sexist.

I believe that the problem will be the opposite - AI will behave in a manner that will be non-discriminatory, and this will cause widespread anger and unrest. Currently, the CPS sentencing guidelines are almost always applied in the most lenient possible manner. Offences such as assault on emergency worker almost never receive anything other than the minimum allowable penalty, and very often well below that. In the hands of AI, you would see people receiving the statutory penalties commensurate with their offending history and the circumstances, and you'd see a lot more custodial sentences. If you look at the demographics of offenders, that is only going to point one way - you will get a lot more young BAME males in prison, and the rallying cry of the social justice brigade will be that the judiciary has become racist robots. The alternative to this, is of course to program in additional leniency - which then makes a mockery of the sentencing guidelines.

The same would go for certain classes of accused. Currently, women almost never receive the same penalties as men for a similar offence, especially violent offences. This is due to a general judicial leniency and taking into account issues like childcare. AI would have to be programmed to take this kind of thing into account - essentially building systemic bias into the software, rather than it being judicial fiat.

As you go higher up the courts, the idea of innovation and landmark decisions would disappear, or be largely discredited. If the Master of the Rolls, or the President of the Family Division is basically Akinator with a law degree, there either won't be any novel case law, or there would be a constant fear of it being a software bug. Take for example Radmacher. Ante-nuptial contracts had never before been given weight - why has this decision gone this way? Are "judges" suddenly making a general move toward this, or is it a software problem? You can't ask an AI for a rationale - it will just regurgitate what has been fed in.

Hi Trinculo

Thank you for the really interesting response there.

What you've done as well is really critiqued not just the notion of AI as judge, but people as judges using a jurisprudence focus. What you've highlighted there is an almost naturalist and realist type critique/observation. The notions of real rules and paper rules conflicting and something above purely positivist statements of law.

I guess as well a question could be, how does the AI work out the interpretation of a statutory provision? Or keep the law updated (you highlighted Radmacher v Granatino)? Courts are often tasked with attributing meaning to vague concepts like 'dishonesty' and 'serious harm', if there is no existing precedent, do you think an AI could ever be aware enough to create a meaning? Or update a meaning? Would you trust it to? Do you think we would be back to your idea of some sort of judicial programming body in charge of updating algorithms etc?

You make an interesting point about appeals as well. Do you think an AI can err in law?

You spoke extensively about sentencing. If you take the custody threshold, for example, how do you think an AI might work out whether it has been passed? The same for the community order threshold?

What do you think of the current state of sentencing?

So I'm not an expect in the law by a long shot at all.

However if you gave me a choice between an AI Jury or Human Jury - I would actually choose the AI jury. Human beings are biased, emotional and easy to manipulate. I've watched so many documentaries of people who were innocent but ended up in prison because the prosecutor has planted an idea. It also works the opposite way - there was a famous case a few years ago of a guy named Richard Scrushy. It was clear that he had committed what he was accused of (fraud) but his lawyers got him out, he pretended to be a Christian, when to "black churches" and during the trial his lawyers made it all about the "American Dream". The jury didn't understand the prosecutors argument as it was full of financial jargon. AI presumably would look at the facts and come to a conclusion.

I think where I would have a human is definitely the judge as they can look at background of the person who committed the crime, whey they committed the crime etc

So I'm not an expect in the law by a long shot at all.

However if you gave me a choice between an AI Jury or Human Jury - I would actually choose the AI jury. Human beings are biased, emotional and easy to manipulate. I've watched so many documentaries of people who were innocent but ended up in prison because the prosecutor has planted an idea. It also works the opposite way - there was a famous case a few years ago of a guy named Richard Scrushy. It was clear that he had committed what he was accused of (fraud) but his lawyers got him out, he pretended to be a Christian, when to "black churches" and during the trial his lawyers made it all about the "American Dream". The jury didn't understand the prosecutors argument as it was full of financial jargon. AI presumably would look at the facts and come to a conclusion.

I think where I would have a human is definitely the judge as they can look at background of the person who committed the crime, whey they committed the crime etc

If COVID and the past 18-months has shown us anything, it is that the work of lawyers can still take place remotely. Whether this is through online courts or completion of legal transactions at a distance or the provision of legal advice.

But take this one step further. All the above examples require lawyer interaction and engagement. Is that needed?

Artificial intelligence is designed to replicate human thinking and acting. It can replicate a human to notice patterns in data. Artificial intelligence can replicate a human in terms of making judgments on a person including whether a person’s smile is ‘real’ or ‘false’.

Can you foresee a legal system where no human interaction is needed – and so human lawyers are replaced rather than their work replicated? We can pay for goods at a self-service check-out. We can order a taxi via an app and in the not too distant future are likely to have a driver-less car pick us up. Adverts that we see on social media are targeted at us by our viewing habits and the profile that has been built up about us. All require no other human interaction.

Why can’t this extend to the legal sector? Can you see a future where a court case is determined by AI examining the papers and coming to a decision? Would this remove all the human bias that may exist within a jury? What implications are there for someone who was perhaps charged with theft or a terrorism offence to attend court for their case to be considered by AI? Is this desirable?

Think ahead and put your thoughts below – there are no right or wrong answers – let’s dream about what future law and the role of lawyers and judges may look like.

This is really interesting, @kwame88. Here's a question though...Is AI only as good as the information that is fed into it? With a jury trial, yes you would certainly get individual biases; that is the nature of it. But would an AI jury potentially share the bias of the creator of the AI? A human jury can check and balance any biases within then as part of their decision making process. How could an AI jury do this?

AI is built on hard facts, systemising processes that can be put into formulas, human judgement is based on experience and logical deductions a lot which would be criminal and unethical to leave to a robot. ive unfortunately had to take someone to court and the discussions weren't over hard facts and numbers, it was about the most logical reasoning based on what has been presented.

perhaps the more mundane work like putting together paperwork for a will or buying a house can be replicated, but law involving disputes, not so much.

AI by machines is intrinsically programmed with one thing in mind - structure, logic, objectivity, problem-solving. Humans rely heavily on true experience. Can these machines really have experiences in the same way? Or is the one objective in the way in which they are programmed merely to retain information that is not yet stored in their database? I would argue that AI would probably overlook the subjective nature of humans that is so necessary in the field of law. It could be limiting. Every case is different and these machines would have to be programmed very well in order to make a well-discerned and fair judgement. There's also the issue of human error vs mechanical error.

But then again, maybe subjectivity is where human lawyers fall short? Maybe we are too emotional and not rational enough?Personally I think human affairs need to be dealt with by humans. It's an interesting topic. I don't know enough about it. This is just where my thinking has led me so far.

I wouldn't be so sure. AIs don't simply look at facts, and particularly in legal cases, facts may require interpretation or contextual assessment. They instead abide by algorithms (sometimes yielded through black box approaches) to enable them to make decisions. And they can be very susceptible to manipulation themselves, if you can figure out what makes them tick.