A recent speech by Sir Andrew McFarlane, the President of the Family Division, England’s most senior family court judge, announced a project to gather data on the financial circumstances leading to family court financial settlements. The English family justice system, as with many justice systems around the world, is keen to explore every opportunity to use AI and algorithms. But it is continually held back by the lack of data. This announcement bodes well for the future. The use of AI in ascertaining and/or facilitating fair financial outcomes on divorce is possible and vital, provided there is the data.
England and Wales, along with most other common law jurisdictions, has a discretionary system for the financial outcomes on divorce. There is significant latitude in order to come to a fair and just outcome according to the particular circumstances of the family. As families are often so very different, with just a few facts potentially changing the fair outcome significantly, it is very difficult to predict outcomes with certainty and confidence. Experienced family lawyers, knowing particular outcomes in reported decisions and with some guidance from the higher courts, are often able to give a fairly good prediction of the outcome. But it is far from ideal.
Inevitably family lawyers have limitations in the number of reported cases they can read, assimilate, remember and analyse. (Moreover these reported decisions tend to be high conflict or quite distinctive facts: a reason they are at appeal or High Court level. They are not middle-income “normal” cases). This is where artificial intelligence can provide solutions, as it has in many other areas of life and will do so increasingly in the future. Analysing a huge amount of data, here being the very many family law cases and the outcomes in those cases, allows therefore predicting with confidence what is likely to be the outcome in any particular set of circumstances. This is something which many of us have hoped could happen over many years. With the development of artificial intelligence, it seemed it was getting close to hand.
Except there was one missing ingredient. The data for the analysis. Court services, here and abroad, have not committed resources to storing details of the financial circumstances in respect of every financial order made. The details of the financial order had not been broken down into constituent parts. Put simply, the big data was not there. So the analysis underpinning AI could not happen, and the predictive outcome was impossible.
I remember in the early 1990s what was then the SFLA, representing family law solicitors, tried to gather this information to carry out this analysis. Well before AI and probably before any meaningful computer usage. It didn’t happen unfortunately and very little has been tried subsequently.
Then in early April 2019, the recently appointed President of the Family Division announced a project to bring together big data and AI in family justice. The digital champagne corks were popping
In spring 2018, I was invited by the Rowntree Foundation to speak on digital technology in the family justice system. Coming to the podium I found sitting in the fourth row Sir James Munby, the then President of the Family Division, Sir Andrew McFarlane, the now recently installed present President and Sir Ernest Ryder, the judge responsible for modernising the justice system. If adrenaline was not flowing on arriving at the podium, it was immediately at that point. In the speech I said as follows.
It is the realm of finance resolution where there is the greatest potential for improvement of resolution through digital interventions. For many civil law jurisdictions, there is relatively little in court adjudications regarding the marital capital. It is only issues of any form of equitable redistribution and maintenance.
In the common-law process, with significant discretion and flexibility, there is far greater difficulty in systematisation. This in turn makes it more difficult to introduce digital technologies in the actual resolution process. Difficult but not impossible.
For many years, some of us have been arguing for a form of equation, formula or at least flowchart for the resolution of financial matters on divorce in a discretionary system. England is in a far better position now than at any time over the past couple of decades with its categorisation of assets and equality of sharing subject to needs. But still there is the issue of needs and when it is appropriate to depart from equal sharing.
Artificial intelligence has had far less application in family law than in many other areas of law. But it can (or could) answer the continued protest of common lawyers that the discretionary fairness exercise of the courts can never be distilled in any digital way. This protest has nullified and stultified opportunities to make good digital progress. But the Law Commission (February 2015: Matrimonial Property, Needs and Agreements) recommended greater use of formula for resolving these matters (Para 3.159):
“We recommend that Government support the formation of a working group, to be convened once suitable empirical data become available, to work on the possible development of a formula to generate ranges of outcomes for spousal support.”
Without creating unnecessarily high hopes, AI can number crunch and fact crunch together all reported decisions and settled cases and the wide variety of factual situations occurring in family law cases to produce either the outcome as will have occurred across a huge variety of cases or at least produce the starting point.
The problem is not the ability of AI. It is the data. With restricted funding over recent years, most family justice systems have not had the ability to carry out appropriate data analysis. The huge task ahead would be categorising the facts of reported decisions at all levels of family courts, and fundamentally unreported cases which have settled without hearings, in order therefore to avoid undue concentration on the high conflict and/or big money cases which go to trial. It’s only once this task has been satisfactorily undertaken could the AI process go ahead. For many involved in pressing forward digital technology and hoping to see a better system, there is real dread that the process will not go forward because of the lack of data rather than the lack of technology or initiatives or the enthusiasm of practitioners.
If this AI investigation and analysis were possible, then for example, if before commencement of proceedings a party was aware through the use of AI data of a huge range of national cases (and also perhaps relevant international cases) that on identical, or at least very similar, facts there was a particular outcome in at least 85% of those cases then for most sensible people this would drive towards settlement and not litigation.
Unfortunately, the use of technology in getting to a resolution has not had a good history in English family law because of the Child Support Agency. Launched as long ago as 1993, it has had constant relaunches, new incarnations, marketing drives and production of ancillary services. But in essence the formula has been flawed from the outset and never fully accepted by the profession and other advisers. It has been badly funded with inadequate powers in those cases which are hard to enforce. The arrears have been staggeringly high. It has probably single-handedly put back commitment to digital means of resolving family cases. The easy response of the English Luddite, seeing the new machinery of this digital legal revolution in family justice, has been to refer to the CSA.
But this historic failure, which started badly with a digital disconnect between government tax records and child maintenance calculations and has not really much improved, must not stand in the way of moving to reform the law of financial outcomes on relationship breakdown. The law must be adapted to digital processes. The law must not be so sophisticated and keen to produce a tailor-made outcome for every individual case as to make it so expensive, so complex and so hard to predict as to put off both lay parties and digital programmers. Of course, IT is the servant of the law but the law must adapt to work best with IT. This is not a challenge to fairness. It is a challenge to produce an accessible justice system.
The central importance was having the data of the financial circumstances giving rise to the financial order or settlement and the terms of that outcome. Once these were available and could be brought together, AI should be able to indicate with a fair degree of confidence what would happen in any particular set of circumstances and thereby directly encourage settlement and avoid unnecessary litigation. Of course, there would be always the very distinctive circumstances where the common-law discretionary opportunities would be most needed. However, for the vast majority of cases, this reliable AI guidance would be sufficient.
Whether as a result of my speech or otherwise, Sir Andrew McFarlane on becoming President has quickly tackled this issue.
On 5 April 2019, Andrew McFarlane gave the keynote speech at the resolution conference. He said as follows
“For as long as I can remember practitioners in ordinary cases, where big money does not even feature in the dreams of the parties, have cried out for authoritative guidance as to the general approach, or even the going rate, applicable in such cases. Active consideration is now being given to harnessing the new computerised process and combining this with a revised Form D81 (the Form D81 Fan Club is, I understand, a very small group!) so that, at the end of every single case, the system will have basic information as to the key financial components and facts, together with the outcome of the proceedings. The plan is for researchers to harness this substantial body of data in order to produce schedules or tables identifying the preponderant outcome in typical cases across a range of set variables. The goal is to provide family lawyers with a resource similar to that enjoyed by personal injury litigators when assessing the quantum of compensation following personal injury.
This is a most excellent development. It has been the hope of a small group of practitioners over several decades. With modest and proportionate beginnings, it should quickly show its worth and warrant the far greater collection of big data which family justice systems need.
The D81 is the form completed by each party when asking for a financial consent order. It sets out in summary the financial and other relevant circumstances so that a judge can be satisfied the intended order is fair in law: judges do not just rubberstamp consent order applications. It is a short but fairly complex document to complete. To make it easier and quicker to distil into data fields, D81 will need to be dramatically redesigned. Specifically, it will have reference to “needs”, a vital component in English financial orders. There may be other information needed for this AI usage which is not presently needed for judicial consideration. So the D81 may need more information. This may be governed partly by the AI practitioners.
Equally the final court orders may need to be restyled to make them easier on big data fields collection. Huge strides have been made with the Family Court Standardisation Programme, overseen by Sir James Munby and by Mostyn J, whereby orders should now be of a standard form. But perhaps it should go further. Inevitably much of the content is very technical, to make sure the parties know what should happen and what has been agreed with legal precision. Nevertheless, in conjunction with this important project, should there be a move towards a new way of expressing orders with perhaps executive summaries before naturally moving into the detail. This would then help data collection hugely and thereby help greater facilitation of AI.
Lawyers often overestimate the importance of sophisticated, tailor-made outcomes. Most clients are far keener to have an early, quick, cost-effective settlement as long as they know it is within a range of what the law may regard as fair. This is what AI can provide if AI has the data.
It should help many cases settle much earlier and without extensive litigation. It will certainly help mediators. It may help judges. Any resource which gives reliable predictions on what is a fair outcome must be welcome.
There is a lot of work needed between digital programmers, family lawyers and family policy advisers. But crucially the work is now starting.
Sir Andrew is to be hugely congratulated and applauded for this initiative, which deserves full support from family lawyers and from digital engineers working with this available data. It is a small step towards a significant transformation of family justice.