Maybe this is just a pet peeve of mine. And maybe it's a pet peeve because I'm a lawyer and am looking forward to a long, fruitful career as a lawyer.
But otherwise very smart people studying and developing Artificial Intelligence keep saying lawyers are one of the most obvious professions to be replaced by robots. Yet they continue to demonstrate that they don't understand what lawyers actually do.
This is from a May, 2015 article that appeared on pbs.org:
Attorneys:
Though it’s unlikely we’ll see robots litigating in courtrooms any time soon, Ford says that some highly billable work normally reserved for seasoned attorneys is in the process of being automated.“We are already seeing an impact in fields like law, with entry level and paralegal jobs which involve document review. It used to be a manual process. They had to read through documents. Now that’s done algorithmically using artificial intelligence.”
“There’s a new emerging technology called quantitative legal prediction. It turns out that experienced lawyers often add a lot of value by making predictions. They’ll do things like tell you what is the likelihood you’re going to win a case, or that the case will be overturned on appeal, for example. It generally takes a lot of judgement and experience to make those kinds of predictions, but these algorithms can actually out-perform even the most experienced lawyers by just looking at lots and lots of data.”
Where to begin with the wrongness of these statements?
1. The expert talks about replacing "paralegal" jobs. But paralegals aren't lawyers.
2. The expert talks about document review. Really? If you're a party in a multi-million dollar lawsuit that involves intensive document review, where there may be one document, in a stack of 20,000, that blows the case wide open, are you really going to rely on a computer to find it? Do you trust a computer to accurately read all text on documents, including handwritten notes? I have the latest Adobe Acrobat software, and it struggles to read things.
Oh, and document review is not done by "seasoned attorneys". It is done by newbies, which the next sentence in that paragraph confirms. In fact, these days it is done more and more on a per diem, as-needed, basis, not by full-time employed lawyers at firms. And that is maybe 0.05% of the legal profession, if that.
3. "Quantitative legal prediction" and "the likelihood you're going to win a case". What does that even mean? What is "winning a case"? Is it winning a case on a motion for summary judgment? Is it getting a case dismissed on a pre-answer motion to dismiss? What if you win before the lower court but lose before the appeals court? What if you lose before the lower court but win on appeal? Appeals courts say it's hard to get them to reverse a lower court, you better make a strong showing, but I've done it. What would a computer say about that?
And what if you get a $10 million jury verdict after a trial in a personal injury case (which could happen in the Bronx). On appeal, you might get nothing because the appeals court could say the jury was wrong! Or the appeals court can knock the award down to $50,000. Which one of those outcomes is a "win" (if any)?
Let's say your case goes to trial. One thing jurors decide is the credibility of witnesses who testify at trial. Can robots predict how a jury will judge a witness's credibility? Of course not. Nobody knows beforehand. But the whole case could turn on one witness's credibility. For instance, in criminal cases, many people have been wrongfully convicted of crimes, released only after decades in prison.
Some (civil) defendants are caught "dead to rights", but choose to fight anyway, to the bitter end, for whatever reason. I've seen it many times. It just means more and more costs and expenses on both sides of a lawsuit. A robot's prediction about winning a case means nothing in that situation.
And by the way, before you can even make a prediction about winning a case, you have to go through the disclosure of evidence, called the discovery process: lawsuits, answers, notices for discovery and inspection, notices to admit, subpoenas, interrogatories, depositions, motions to compel, motions to preclude, motions for protective orders, court conferences along the way, etc. Are robots going to do any of that? No. But the AI expert does not include any of that in his statements--because he doesn't know.
This nonsense about "quantitative legal prediction" is meaningless in the real world. It has no practical application and shows these AI experts have no idea what lawyers actually do or how the legal system actually works.
But otherwise very smart people studying and developing Artificial Intelligence keep saying lawyers are one of the most obvious professions to be replaced by robots. Yet they continue to demonstrate that they don't understand what lawyers actually do.
This is from a May, 2015 article that appeared on pbs.org:
Attorneys:
Though it’s unlikely we’ll see robots litigating in courtrooms any time soon, Ford says that some highly billable work normally reserved for seasoned attorneys is in the process of being automated.“We are already seeing an impact in fields like law, with entry level and paralegal jobs which involve document review. It used to be a manual process. They had to read through documents. Now that’s done algorithmically using artificial intelligence.”
“There’s a new emerging technology called quantitative legal prediction. It turns out that experienced lawyers often add a lot of value by making predictions. They’ll do things like tell you what is the likelihood you’re going to win a case, or that the case will be overturned on appeal, for example. It generally takes a lot of judgement and experience to make those kinds of predictions, but these algorithms can actually out-perform even the most experienced lawyers by just looking at lots and lots of data.”
Where to begin with the wrongness of these statements?
1. The expert talks about replacing "paralegal" jobs. But paralegals aren't lawyers.
2. The expert talks about document review. Really? If you're a party in a multi-million dollar lawsuit that involves intensive document review, where there may be one document, in a stack of 20,000, that blows the case wide open, are you really going to rely on a computer to find it? Do you trust a computer to accurately read all text on documents, including handwritten notes? I have the latest Adobe Acrobat software, and it struggles to read things.
Oh, and document review is not done by "seasoned attorneys". It is done by newbies, which the next sentence in that paragraph confirms. In fact, these days it is done more and more on a per diem, as-needed, basis, not by full-time employed lawyers at firms. And that is maybe 0.05% of the legal profession, if that.
3. "Quantitative legal prediction" and "the likelihood you're going to win a case". What does that even mean? What is "winning a case"? Is it winning a case on a motion for summary judgment? Is it getting a case dismissed on a pre-answer motion to dismiss? What if you win before the lower court but lose before the appeals court? What if you lose before the lower court but win on appeal? Appeals courts say it's hard to get them to reverse a lower court, you better make a strong showing, but I've done it. What would a computer say about that?
And what if you get a $10 million jury verdict after a trial in a personal injury case (which could happen in the Bronx). On appeal, you might get nothing because the appeals court could say the jury was wrong! Or the appeals court can knock the award down to $50,000. Which one of those outcomes is a "win" (if any)?
Let's say your case goes to trial. One thing jurors decide is the credibility of witnesses who testify at trial. Can robots predict how a jury will judge a witness's credibility? Of course not. Nobody knows beforehand. But the whole case could turn on one witness's credibility. For instance, in criminal cases, many people have been wrongfully convicted of crimes, released only after decades in prison.
Some (civil) defendants are caught "dead to rights", but choose to fight anyway, to the bitter end, for whatever reason. I've seen it many times. It just means more and more costs and expenses on both sides of a lawsuit. A robot's prediction about winning a case means nothing in that situation.
And by the way, before you can even make a prediction about winning a case, you have to go through the disclosure of evidence, called the discovery process: lawsuits, answers, notices for discovery and inspection, notices to admit, subpoenas, interrogatories, depositions, motions to compel, motions to preclude, motions for protective orders, court conferences along the way, etc. Are robots going to do any of that? No. But the AI expert does not include any of that in his statements--because he doesn't know.
This nonsense about "quantitative legal prediction" is meaningless in the real world. It has no practical application and shows these AI experts have no idea what lawyers actually do or how the legal system actually works.
Comments
Post a Comment