Science has a better chance of predicting the outcome of litigation, they said, and one of the most interesting places to glean data is from legal pleadings.
In common law countries, litigation data analysis of both precedents and case law can mine the text for words that have predictive value.
Pleadings are private
However, while court judgments are available electronically in the Irish courts system, pleadings are not, and remain private to the parties involved in the litigation, Larry Fenelon of Leman noted.
Litigation is often a hard slog, he observed, and usually an enormous distraction to the management of a company.
As well as money spent on lawyers, there can also be negative press.
He suggested that a general counsel who could produce statistical analysis or a predictive theory of litigation prospects, based on poor corporate governance in an organisation, could be highly influential.
Preventing the bad stuff
“I may sound like a turkey voting for Christmas, but the real value I see in litigation-data analytics is the preventative value.
“The true calling of lawyers is preventing the bad stuff from happening in the first place. That’s where you gain a huge degree of trust and value among your clients,” Fenelon said.
It is an enormous practical task to set up a statistical modelling unit in any legal organisation, unless professional data scientists are brought in, Fenelon observed.
Prof Daniel Lee explained that text mining has now expanded from numerical data to words.
However, data scientists spend most of their time in cleaning data, as well as in joining data from different sources, by finding a common variable.
Cleaning data is time-consuming, and having structured inputs from the start is less risky and best practice, he said.
Prof Lee said that missing data values cannot be intuited blindly, because they are often missing for a reason.
“I’m not a lawyer, and I wouldn’t know why there is a missing value. Is it a typo or an error in the collection process, or maybe somebody didn’t want to put it in, for a reason?”
This is why data analytics teamwork is important, and must include both those with legal domain knowledge, aligned with the technically-skilled.
Prof Forgues said that fewer lawsuits mean fewer business interruptions and more efficiency.
But, once litigation is underway, the text of documents, evidence, and correspondence from opposing counsel, can be mined for predictive value as to the success of the lawsuit itself, as well as of certain motions along the way.
This is useful in establishing the probability of success, and in decision-making on where a litigation budget should be spent.
Once litigation is underway, data analytics also makes the process more efficient, with less time spent on less probable strategies, and fewer back-and-forth discussions about damages.
“You will be able to make litigation decisions with confidence, because you won’t spend money on strategies that have less mathematical probability [of success],” she said.
“The whole thing will be streamlined and, in the long run, this is a money-saver if you can prevent lawsuits and very large-scale costs.
Larry Fenelon asked how legal organisations could ensure that their data from physical documents was clean, properly digitised, and searchable, since the quality of input data is crucial.
There are many different ways to import datasets into data-mining software, whether from an Excel spreadsheet, a company database, or a web page, he was told.
The academics said that lawyers would need to learn how to use ‘R’ – a programming language and free software environment for statistical computing, in the same way that they learned how to use spreadsheets.
“It’s free, and it’s really easy,” said Prof Forgues. “It’s more of a scripting, as opposed to a computer-programming language, and it’s really user-friendly.
“Even lawyers who are not data scientists can learn it pretty quickly,” she said.
While general counsel should ideally build their own tools, using in-house data analytics, law firms could use external, commercially-available tools.
When a lawsuit is filed, the text can be scraped from the complaint page and used as a predictor for the legal outcome.
Firms such as LexMachina and Westlaw Edge focus on litigation-management analytics and strategy.
Using machine learning, each word becomes a binary variable, taking either zero or one, and that becomes a predictor.
“We are getting better at it,” said Prof Lee, though he accepted that words can be used in different contexts.
Where there is missing data, this is intuited – though this should be handled with care, the academic explained.
In the US, data-analytic capacity is now extending to all litigation meetings, with respect to damages requests and awards.
External data about particular companies and certain industries can also be mined, to indicate predictors of whether any particular firm is going to be sued.
Economic, industry and demographic data are also on tap.
‘Hold or fold?’
Larry Fenelon of Leman pondered whether the number crunching and data analytics would suggest whether one should ‘hold or fold’ in terms of settlement, or going to hearing.
Fenelon commented that he already frequently intuits from the language in a legal letter as to whether a case will succeed or fail, on a ‘finger-in-the-air’ level.
“When someone talks about their human rights in a breach-of-contract case, I’m always a bit more suspicious,” he said.
Statistical modelling will eventually have predictive mathematical value, the academics responded.
Data gleaned from employee performance reviews currently had good predictive value about employment lawsuits, the symposium heard.
Fenelon said that data predictions of whether employees might leave an organisation could be very useful, since it is so much harder to replace than to recruit a worker.
A large publicly-traded company will have more data available than a small private company, however.
It may take some years to get value from the data that is gathered, Prof Lee said, and collecting data on competitors was also very important.
Getting sued more often
“If you are getting sued more often than your comparables, we can use the data to investigate that question,” he said.
Prof Forgues cited the example of a poor middle manager in the healthcare sector costing one hundred lawsuits over ten years.
“If you’ve got a good CEO, that will actually subtract a certain number of lawsuits, so it’s really important information.”
For lawyers to put machine-learning technology into practice, they need to start collecting data and building the models, putting everything in searchable forms, Prof Forgues said.
Simple mining of this text data will throw up interesting patterns, even without the precision of statistical modelling, she said.
She added that a smaller family law firm in the US had used data analytics to figure out which of their clients were less likely to pay their bills, by collecting financial, text and demographic data.
The data showed that the punters less likely to pay their legal bills were all using a gym that was running an ad for a local legal malpractice lawyer.
“That’s a small-use case and it picked up on something that the human brain wouldn’t normally pick up on – people who visit this gym are less likely to be paying clients.
“If you’re a small practice and every dollar counts, that’s a way to put analytics to good use,” Prof Forgues said.