Use of AI predictive tactics could accelerate the reduction in cases going to trial, a trend that has been increasing for years. Conversely, it could increase trial frequency as some lawyers may view trial as more predictable and thus, advantageous. In any event, a significant portion of pretrial practice is likely to be cut out to make way for predictive analysis.
In the face of successful predictive modeling, a reduction in trial is probably the more likely outcome. Currently the federal numbers say that only around one percent of civil cases and two percent of criminal cases make it to a trial decision. Extensive pretrial processes are partly responsible for this. The intended purpose of pretrial is multifold—it is there to force the parties to interact, to facilitate proper discovery, and to ensure that all other options are exhausted before trial is resorted to. See Travis Mills Corp. v. Square D. Co., 67 F.R.D. 22, 26 (E.D. Pa. 1975), see also 207 Pa. Code Part IV, Art II, Ch. 4, Rule 421. Dedicating a significant portion of the pretrial period to predict juries, judges, mediators, and arbitrators can shift the purpose of pretrial activity from fair communications between the parties more toward the purpose of using technology to predict the future. AI infecting litigation in this way may greatly disrupt the ends of equity that pretrial procedure was created to achieve; such a pathogenic effect would prove detrimental.
The harm caused by a finite number of trials could be serious and, in the most dramatic of possibilities, may direct pretrial practice to a place where trial is never expected and rarely prepared for. As the asymptotic number of trials creeps closer and closer to 0, they could become a truly anomalous occurrence. An asymptote is a curve that infinitely approaches a point (zero), but never touches it. On the opposite end, AI predictions could possibly highlight advantages of taking cases to trial and create a larger affinity for it among attorneys. Again, however, this result is not likely. Not only might predictive AI smother out the already diminishing flames of trial, but it could also be protected and incubated through federal and state rules of civil procedure.
Federal Rule of Civil Procedure 26 would be responsible for ramparting AI’s predictions as attorney work product. Once the anticipation of litigation begins, the thought processes and strategies of lawyers are afforded the protection of “work product.” Fed. R. Civ. P. 26(b)(3). By the letter of the law, work-product protection is there to safeguard, among other things, the impressions and theories of attorneys working on the case. McClanahan v. United States DOJ, 204 F. Supp. 3d 30, 51 (D.D.C. 2016). Predictive impressions of future juries or tribunals certainly qualify as such, leading to the conclusion that any AI analytics would not only be feasible to engage in, but would also be afforded civil procedural protections.
Even without trial, predictive analytics can permeate through to other tribunals or decision makers like in mediation or arbitration; there are already private AI models that claim to predict judges and neutral decision makers’ decisions up to nearly 90 percent accuracy. AI is everywhere, and its predictive and generative capabilities are not going unnoticed.
The trend of bringing cases to trial is already a downward sloping line. If predictive AI were to prove effective and subsequently, widely infect pretrial practice, it is possible that this trend would inch closer and closer to 0 as an asymptotic curve, reserving juries for the most novel of questions and the wealthiest of litigants.
