There is a lot of activity happening under the umbrella of artificial intelligence. In this country and in the field of access to justice, much of it is, however prospective – promised rather than realised. In the US, matters are further ahead and there are interesting developments like the Learned Hands gamification of problem classification that may well take things forward in a concrete way.
In England and Wales, we are particularly blessed with committees and competitive grant schemes. The Judiciary has just appointed an advisory committee chaired by the omni-present Richard Susskind. The Law Society, the professional body of solicitors allegedly a little miffed at their members’ widespread absence from the judicial body, has set up a public policy commission chaired by its President Christina Blacklaws. Slightly bizarrely, the Department of Business, Energy and Industrial Strategy has given the Solicitors Regulation Authority £700,000 to further kickstart the growth of AI in the legal profession and examine the implications. The SRA has subcontracted with Nesta (once more understandably known as the National Endowment for Science, Technology and the Arts) actually to do the business.
And even more money is coming from a joint Department of Business and Digital, Culture, Media and Sport Next Generation Services Industrial Strategy Challenge Fund. This apparently amounts to £6.4m given to ’18 legal artificial intelligence and data analytics projects’. Much has gone to commercial or academic recipients but it also ‘included £262,000 for consumer website and forum Legal Beagles and IBM, working together on ways of using AI to “predict best routes for consumers to find solutions to legal issues” and “locate legal knowledge faster, identify new patterns and trends, whilst at the same time helping consumers with their legal issues”. In additoin, ’a project on affordable legal advice, involving the Royal Courts of Justice, Solicitors Pro Bono Group and Islington Citizens Advice Bureau among others, was awarded £182,000.’
The IBM/LegalBeagles tie up sounds very close to an interesting joint project between Suffolk and Stanford Law Schools in the US. This combines an attempt to classify types of legal problems; to use questions from ordinary people to test application of those categories; and to prepare the ground for automated referral and resolution. Its most visible product is the Learned Hands gamification. You can listen to a Law Talk Network podcast or read various articles on this project. It involves an alliance between two of the big beasts of the access to justice tech scene in the US, Suffolk’s David Colarusso and Stanford’s Margaret Hagan. Interestingly, the idea for the collaboration appears to have arisen from a further example of cross-fertilisation — the two principals came up with the idea at the Legal Services Corporation’s Technical Initiatives Grants conference.
This is Stanford’s description of Learned Hands: it ‘is a game in which you spot possible legal issues in real people’s stories about their problems. You read the stories, and then say whether you see a certain legal issue — family law issues, consumer law issues, criminal law issues, etc. The game is also a research project. Each time you play, you are training a machine learning model to be able to spot people’s legal issues. This model will be used to develop access to justice technologies that connect people with public legal help resources. It will help us to make a Rosetta Stone for legal help — linking the legal help guides that courts and legal aid groups offer to the people who are searching for help.’ And this is David Colarusso’s pitch on its value: ‘Something like an automated issue spotter has the potential to improve access to justice simply by making it a little easier to find legal resources. It doesn’t need to answer people’s questions. It just needs to point them in the right direction or bring them to the attention of someone in a position to help. It can get the conversation started by making an educated guess about what someone is looking for and jumping over a few mundane—but often intimidating—first steps.’
Lying behind this project are some earlier pieces of work. Margaret Hagan has been bringing up to date a National Subject Matter Index. ‘Right now, most legal issue taxonomies are created by lawyers or legal website admins, to better present the resources they have so that people can find them. Or, to better encode the tasks they’re working on, for reporting or billing purposes. The goal of the project is to make these professional taxonomies more streamlined — but then to also make them fit the mental models, issues, and phrases that regular people use.’ So, for example, this has a high level category ‘family’ and then drills down to different elements of problem.’
At this point, enter Reddit. This styles itself as the ‘front page of the internet’. You can write in apparently with pretty well any thought that occurs to you. At this moment, the leading post is the unpromising ‘If Jesus was unavailable who would you scream at to take the wheel?’. Suggested answers included Jeremy Clarkson – amplified by a second contribution which said ‘anyone but Hammond’ – which, at least, shows a familiarity with world media. Notwithstanding this, Reddit is also used by people asking serious legal questions and the project has obtained a cache of anonymised questions. The game consists of spotting ‘possible legal issues in real people’s stories about their problems. You read the stories, and then say whether you see a certain legal issue — family law issues, consumer law issues, criminal law issues, etc.’
For the record, I accumulated 450 points playing the game, regrettably short of Margaret Hagan and Jacob B who each have well over 1m. I am ranked 262 and have a quality score of 0. It is quite interesting to do and I recommend signing up. It is not, as far as I could see, particularly jurisdiction specific. The fun way to participate would be with a group where you debate if there are issues of particular types wrapped up in the questions. I was left a bit unsatisfied. Many of the most difficult questions are those which appear to be one thing – say a family issue – but actually contain a contract point. I was also a bit unsure if I was doing it right: there was a bit of a lack of feedback.
It is interesting to compare the Learned Hands game with the live traffic feed that you can get off the Citizens Advice site in this country. I have been watching this for the last 10 minutes: it summarises the questions that those on the website are asking. There were 1445 in the last minute as I write. Issues might be being missed but actually the summaries both of queries direct to the site and redirected from google – which, I assume, owe nothing to artificial intelligence – are coherent enough to allow fairly accurate referral. There is the occasional odd ball. Heaven knows what ‘building a bridge’ referred to (I would guess relationships) but the next five questions were summarised coherently enough at least for referral ‘criminal record, pip [a disability benefit], sick note, divorce, grievance [I assume an employment issue]’.
The important lesson is that all these initiatives need to be followed, analysed and compared. The process of automated referral and resolution raises common questions for all jurisdictions, however the detail might differ. And you do have to believe that artificial intelligence has a role to play – though exactly where it is best deployed is to be seen.