A team of computer science researchers has developed an AI system that can predict whether or not a European human rights complaint would be considered legitimate by judges, with a 79% accuracy level.
The researchers from University College London (UCL) and the universities of Sheffield and Pennsylvania, have created the system for cases heard at the European Court of Human Rights (ECHR).
One of the key researchers, Dr Nikolaos Aletras, who led the study at UCL, said of using AI to predict cases: ‘What we think is [that lawyers will] find it useful for rapidly identifying patterns in cases that lead to certain outcomes.’
‘It could be a valuable tool for highlighting which cases are most likely to be violations of the European Convention on Human Rights,’ he added.
The choice of the ECHR was perhaps driven by the fact the court sees huge numbers of cases that the court’s legal staff had to consider. For example, in 2015, the court had to process 40,650 applications for a hearing, while in 2014 it processed 56,200 cases.
At the same time only 2,441 applications to the ECHR received a judgment of some form in 2015. But, a significant proportion of these applications were joined, with the result that the number of judgments actually delivered was 823. The vast majority of applications to the ECHR therefore appear to be struck out or classified as inadmissible to the court.
As can be seen, the court is flooded with demands for a hearing, many of which are not likely to win. In fact, it would appear the court spends a huge amount of time just trying to decide which applications have validity, at least as far as the ECHR can hear such cases.
The AI research team therefore created a type of ‘triage’ system to cut through the cases and give lawyers and judges some a priori indication of their validity. It did not however provide a prediction on specific rulings, but rather whether the case could proceed at the ECHR.
Of course, this does not remove the need for lawyers, nor even if a claimant has received an indication via the AI predictive system will the claimant necessarily wish to give up their case – though their lawyers could then perhaps decide to try a different approach.
According to Motherboard the team first trained a Natural Language Processing (NLP) neural network on a database of court decisions, which contained the facts about the cases, the circumstances surrounding then, the applicable laws and details about the applicants such as country of origin.
Next, the team fed the program human rights court decisions that it had never seen before and asked it to guess the judge’s ruling, based on the constituent parts of the court’s decision filing. They found it was surprisingly accurate when using the model their AI had developed.
The team has also shown in this example how a combination of NLP and machine learning can extract extremely useful information from legal documents, in this case in terms of litigation outcomes.
There is a growing interest in the legal tech field in developing litigation prediction software. For example, Premonition studies which lawyers perform best in court, so clients can make a better choice on who to instruct.
Meanwhile, Lex Machina, which is owned by Lexis Nexis, mines litigation data, revealing insights about judges, lawyers, parties, and the subjects of the cases themselves, culled from millions of pages of litigation information. The idea is that users can form a more accurate predictive picture of a case.
It therefore seems highly likely that applying AI to legal case prediction will become a growing area in the years to come.