Recent years have seen the launch of several so-called “data analytics” tools, designed to help litigators to predict the outcome of their disputes. I think the first one I read about was “Lex Machina”, launched in around 2010 as a spin-off from a Stanford Law School public interest project. At that time, the product was being marketed as “moneyball for litigators”. It worked by trawling publicly available data (mainly, I think, published judgments and decisions), to produce analyses of the relative strengths or weaknesses of a particular claim or defence. The tool analysed a number of factors, but the key selling point seemed to be the ability to predict how a particular judge might decide any given dispute.
Since then, our friends in the tech world have been busy, and there are now several such products on the market. Each seems to take slightly different angles in their marketing spiel; for example, some focus primarily on advocates’ track records with a view to predicting likely outcomes. Others incorporate a focus on the economic utility of settlement, using “game theory” instead of “decision theory” to predict the likely economic consequences for a business of settling, or failing to settle, a dispute at a given point in time. (“Game theory” has been defined as “the study of mathematical models of conflict and cooperation between intelligent rational decision-makers”: insert your own lawyer joke here).
One thing that all these tools have in common is that they are marketed to third party funders, and indeed it is easy to see that funders must form an important segment of the target market. Third party funding is now well established, and is becoming increasingly significant in the context of international arbitration. Presumably, the funders (whose interest in any given arbitration is purely financial) would prefer to have such tools at their disposal in arbitration as well as litigation.
That raises the question of whether, and if so, how, data analytics can be used in that context. The data upon which the software operates is publicly available. What the tool provides is the ability to analyse, quickly and accurately, the significance of data that is already freely available. As the software develops and becomes more sophisticated, no doubt the accuracy of the predictions and analyses will improve. However, the underlying data must be available to start with. Arbitration is, of course, confidential. That is one of its main selling points. There is no comprehensive public data on individual arbitrators or their awards; the picture is incomplete and patchy. So, against that unpromising background, how can data analytics work in the context of arbitration?
I read towards the end of last year that a new product – Dispute Resolution Data – has teamed up with certain arbitral institutions to provide arbitration-specific data analytics. The product was established by an ex-head of the American Arbitration Association (AAA), and is supported by a number of experts worldwide, including Adrian Winstanley, previously Registrar of the London Court of International Arbitration (LCIA). But, in order to preserve the confidentiality of the underlying arbitrations upon which the data analysis is performed (and upon which the institutions’ success depends), the website confirms that “DRD will not request or publish any information from institutions regarding parties identity, the names of the case, the advocates, the arbitrators or mediators, nor associate institutions’ names with their data.” So, it looks as if (unlike the litigation tools) it won’t be possible to run analyses based on individual arbitrators.
Stepping back for a moment, I don’t think this is such a bad thing, at least from the point of view of the arbitration community as a whole. If data analytics tools could be used to predict an individual arbitrator’s decision in any case, it seems likely that existing patterns of appointments of certain arbitrators to certain types of dispute could be reinforced and perhaps become even more entrenched. That is hard to square with recent initiatives to improve diversity and transparency in arbitral appointments, and to address the perception that arbitration has become a bit of an old boys’ club.
But it is also clear that any product that is founded on partial data can’t be as reliable. So unless a sufficiently large proportion of institutional and ad hoc arbitration could be brought within the scope of data analysis, the tool is only going to give you a fairly broad brush picture. That might still be very useful, of course. But it won’t let you read the arbitrator’s mind.
Of course, the more accurate these tools become, the more difficult questions arise. For example, increased accuracy might mean that parties (or, more pertinently, funders) ask themselves whether it is worth litigating or arbitrating at all. Why go through the pain and cost of a trial or hearing simply to confirm the accuracy of the cyber-prediction? Here, we’re getting close to the notion of disputes (in effect) being determined by artificial intelligence – not all bad, I suppose, and at least it should be cheaper than human intervention. But if nobody will pay to challenge the predictions of the software, this must mean that there is an ever-decreasing pool of current data upon which the software can operate. Does this mean that the tools might become the victims of their own success? And how can the common law continue to develop, in the absence of the “life blood” of court decisions and arbitral awards subject to appeal? Similarly, one also wonders whether there would need to be rules around the disclosure of cyber-predictions to a tribunal. Might an arbitrator feel pressured into agreeing with the analytic tool? After all, the computer has been programmed to “get it right”, free from human error or weakness. There must surely be a possibility of some sort of confirmation bias operating here.
Thankfully, it isn’t quite time to kneel before our robot overlords, but there are certainly some interesting developments to keep an eye on in the context of dispute resolution generally and arbitration in particular.