Mediation has become a popular alternative to expensive legal litigation. Throughout the process, trained independent mediators work to impartially guide parties from the opposing sides of a legal dispute to a mutual agreement.
San Antonio lawyer and mediator Don Philbin wondered if predictive analytics would be effective in steering clients’ next moves in negotiations.
As it turns out, the answer is yes.
Predictive analytics have entered the legal sector to help lawyers craft concession strategies for clients regarding how much to ask for in a monetary settlement at what time during negotiations.
Philbin, who specializes in dispute resolution, likened mediation to the traditional process of buying a car. The back and forth of dollar amounts represented by “bids,” or offers, and “asks,” or demands, in a single mediation session can span seven to eight hours, as both sides typically hesitate to “give in” too soon. Ending the process after only one or two rounds of offers and counteroffers could lead to a party walking away from a mediation session – or a car dealership– feeling they conceded too early and paid too much.
Some seasoned pros will tell you that they know within three hours of the start of a mediation whether a settlement will be reached, Philbin said. “The consensus was that by then, you’d have a pretty good bead on what the final numbers will be.”
This insight drove Philbin to collect qualified data from advocates and mediators in order to test the hypothesis that there is predictability in negotiations. A mock-up of an early version of a software algorithm demonstrated his theory: There is predictability both on the final agreed-upon settlement amount and on how long it takes to get there.
Philbin sought out mediation attorneys, statisticians, and scientists from San Antonio’s Southwest Research Institute (SwRI) to develop intelligent software that “learns” negotiation strategy from the patterns in thousands of litigated cases. SwRI’s Intelligent Systems division built both the software and an app for smartphones.
The Picture it Settled software shows how far apart the two sides are at the start of a negotiation. Once one side makes an offer, the other can counter. That data is then plotted at two converging lines that mediators hope will meet at an agreed-upon final amount at a specific point in time during the mediation session.
[WPGP gif_id=”963504? width=”2048?]
The software plots each party’s opening offer and counteroffer and at what time within a daylong session they were made. Two curving lines in the graph start far apart at the top, but move toward one another and a likely value.
This “hurricane graph” plotting shows how the software can accurately model people’s well-documented behavior of reciprocity, offering dollar amounts in successive rounds of “tit for tat.”
Deep learning is a new area of machine learning research to help move computing closer to its end goal – artificial intelligence. SwRI used a deep-learning algorithm for the software because its artificial neural networks require large amounts of data to train and refine its predictive capabilities. Neural networks analyze negotiators’ behavior in thousands of cases, enabling the software to forecast what opponents will do as the mediation progresses.
“With more data, you are able to generate better predictions,” Philbin said. “SwRI has it set to retrain the algorithms every night based on new data so that the algorithms get smarter over time with better projections.”
The software was first conceptualized in 2011. Not only has it gone through countless iterations, it has gotten more intelligent and better at predicting outcomes in negotiations, Philbin said. Lawyers use the software to reverse-engineer how to get to a target point assuming the negotiations carry on long enough.
“Parties set their target at say [achieving agreement by] 5:30 p.m. with a specific dollar amount in mind, then use the software to reverse-engineer and plot the path to get to that point,” Philbin said. “Using those insights, they can develop concession strategies that increase the likelihood of approaching that point.”
Picture it Settled also helps adjust client expectations as offers and demands begin to inform the projections while negotiations are underway, Philbin explained. With an average of seven rounds in a typical mediation session, unrealistic expectations can make for a protracted struggle rather than a successful concession pattern that increases the odds of a deal.
“The predictive analytics help guide users to what is probable rather than what is possible,” Philbin said.
The pairing of big-data analytics and legal practices has sparked another successful startup operating in San Antonio: Easy Expunctions leverages algorithms to streamline the process of expunging arrest records or criminal court proceedings for an accurate, cost-effective alternative for clients.
“Lawyers use case precedent to help inform outcomes in similar cases,” Philbin said. “Predictive analytics allow lawyers to look beyond possibilities in one case to what happens probabilistically in batches of cases.”
When asked to forecast the impact of big-data analytics on the legal industry, Philbin offered his own prediction.
“I expect there will be a consolidation in the legal technology industry,” Philbin said. “At some point, emerging predictive tools like this will be assembled to create a powerful toolbox for lawyers.”