Why the BBC was wrong about AI and lawyers – and driving instructors
Have you had a play with the BBC’s ‘Will a robot take your job?’ app?
It’s part of a series they ran last week on intelligent machines. The app allows you to input your job, and it will then tell you what the chances are that your job will be carried out in the future by a robot, powered by AI.
It is based on the research of Benedikt Frey and Michael A. Osborne of the Oxford Martin School’s Programme on the Impacts of Future Technology.
Good news for lawyers – according to the app, solicitors and barristers are ‘quite unlikely’ to lose their jobs to robots, 320th most likely out of 366 jobs assessed. Not so good for legal secretaries – they are the 3rd most likely apparently. So, if you’re a lawyer, ‘computer says yes, you are OK. Chill.’
the underlying assumptions for the research are flawed
But it’s not that simple. In my view, the underlying assumptions for the research are flawed, at least as far as lawyers are concerned.
The research took the standard skills needed for roles and tasks for hundreds of jobs from the US data agency O*Net, and categorised them into a number of types: perception and manipulation; creative intelligence; and social intelligence.
It then looked at whether those tasks could be carried out by intelligent machines according to those characteristics, with those tasks which are non-cognitive and routine most likely to be automated. All very logical.
But there are two major problems with this approach, at least as far as lawyers are concerned.
Problem one: the lawyer role appears to have been taken from the sort of courtroom lawyer you see in US TV dramas.
Here are the headings for the tasks for that role:
- Represent clients in court or before government agencies
- Present evidence to defend clients or prosecute defendants in criminal or civil litigation
- Select jurors, argue motions, meet with judges, and question witnesses during the course of a trial
- Study Constitution, statutes, decisions, regulations, and ordinances of quasi-judicial bodies to determine ramifications for cases
- Interpret laws, rulings and regulations for individuals and businesses
- Present and summarize cases to judges and juries
- Analyze the probable outcomes of cases, using knowledge of legal precedents
- Examine legal data to determine advisability of defending or prosecuting lawsuit
- Evaluate findings and develop strategies and arguments in preparation for presentation of cases
There’s very little in here that the average commercial lawyer does regularly. Nothing much about drafting contracts based on standard documents and a clearly defined body of law, sending them to and fro, negotiating and adding in expected amendments and lodging copies with relevant regulatory authorities.
The BBC has made it worse by treating solicitor and barrister roles as exactly the same (both have identical rankings). But for the majority of solicitors, there is much more work that is routine (and non-cognitive) than for advocates.
So that’s problem 1, the narrowly defined lawyer role.
Problem two is more serious: it doesn’t look at outcomes
The assessment looks at the skills needed now, and assumes that tasks and roles will stay the same but just be automated.
It doesn’t look at whether the outcomes from the job will still be the same, nor whether that role is still required.
It’s the equivalent of asking on the advent of the internal combustion engine whether the skills of horse whip makers to make whips could be mechanised by engines, but without asking how far people would still need horses (and therefore whips) at all.
You can see this pretty easily – driving instructors come out at ‘quite unlikely’ to be automated in the BBC’s test, 265th of 366. But if cars become more or less driverless, will we need driving instructors? I don’t think so.
the very outcomes lawyers will need to achieve will change
As I see it, it’s not just that the tasks lawyers perform will change, but the very outcomes lawyers will need to achieve will change.
- In a world where commercial risk is assessable be reference to quantifiable data, why will we need drafting skills to hedge risk in contracts through clever drafting?
- Where outcomes of litigation can be calculated within a range of possibilities, why will we need the skills to persuade through skilful argument?
- If machines can tell us whether someone is likely to be telling the truth or not, do we need the social skills to be able to read other people’s body language?
Ron Friedmann has been talking this week about the work of Ethereum, in developing the blockchain technology underlying Bitcoin, to reduce the “if… then” risk apportionment in contracts into code and then into deliverable outcomes. I don’t pretend to understand it all, but it sounds like the embodiment of whatProfessor Daniel M, Katz Assistant Professor of Law at Michigan State University has been talking about for some time – contracts as code, not as words.
If that comes to pass, then the skills needed for embodying legal and commercial risks into a form that parties to a ‘contract’ relationship can rely on will be very different from what we expect today.
Yes, there will be high value, ground-breaking, business critical work where there is no prior body of data on which to rely, and which can’t be reduced to data. But that won’t make up the majority of commercial lawyers’ days. It doesn’t now.
So, don’t rely on being 320th in line out of 366. I fear life is much more uncomfortable for lawyers than that.
24 Sep 2015