Posted on March 9, 2020 by Joseph Lamport
When it comes to forecasting the future effects of technology, there is a well-recognized maxim known as Amara’s law, named after the technologist Roy Amara. In his words, “[W]e tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” The mind, in other words, quickly leaps to conclusions without regard for the often lugubrious pace of the tech adoption cycle.
Amara’s rule seems particularly apt when it comes to our attempts to forecast the impact of AI/ML technologies in the legal market. So much of our recent anxiety has been driven by overblown fears about imminent job losses, which seemed certain to result from the wave of AI/ML tools and technologies being brought to market. We jumped to the conclusion that since these tools now exist, and are becoming increasingly powerful with every passing day, their impact will be decisive and swift. And yet, week after week, month after month, we have discerned little or no negative impact on employment in the legal market. In fact, legal employment continues to inch up in the face of ceaseless legal tech innovation. This means we are presently very much in the grip of the first half of Amara’s law, coming to a reckoning that we have likely overestimated the short run impact of AI/ML on the legal market.
Having gotten to this point, it’s not hard to see how the second half of Amara’s law will flow naturally from the first. After we overestimate the short-term impact of new technology, we are then inclined to overcompensate as we develop our long-term views. In other words, once the short-term effects we anticipated fail to materialize, it’s easy for us to get lulled back into imagining the status quo will simply persist forever. It’s worth noting that Amara’s law seems to rest on an inherent flaw of human intelligence (otherwise known as leaping to the conclusion) that may be easily corrected or compensated for by more data-driven machine intelligence.
In any case, with or without the assistance of artificial intelligence, we would all be well advised to try and steer a middle course as we make predictions about the future, neither getting carried away by fears about imminent market disruption nor losing sight of the massive disruptions that will eventually come our way.
* * * * * *
Nowhere in the legal market is the impact of AI/ML technology more likely to have dramatic impact (in the long run and maybe even in the medium run) than when it comes to the business of litigation. We can already see the broad outlines of this future taking shape. Take a look, for instance, at our coverage of the new product just released by Casetext which is capable of generating the first draft of a legal brief in an almost fully automated fashion. This new tool, powerful in itself, is built using next-generation AI/ML algorithms, which are much more effective in emulating the thought processes and patterns of legal reasoning. In other words, the AI/ML technology is rapidly advancing and becoming increasingly capable of performing case law analysis, risk assessment and predicting outcomes.
The next sign of the coming disruption in the litigation business is the rapid development of various analytic tools that are already proving very instrumental in helping litigators shape strategy and tactics. Our colleague Brad Blickstein has provided a useful survey of the current crop of these analytic tools and he has also compiled a more in-depth report (available for purchase) that takes a closer look at precisely what these analytic tools are capable of doing, and the various techniques they deploy for extracting data from caselaw databases to assess and predict litigation outcomes.
A number of these analytic tools have been developed by the traditional legal research vendors, such as Westlaw Edge or Lex Machina, now part of LexisNexis. Others come from smaller start-ups, such as Gavelytics, Solomonic and Premonition. For the most part, the focus of these tools today is to provide data-driven insights, which help litigators evaluate the likely outcomes of bringing particular motions in particular jurisdictions, or how an adversary has faired before a particular judge. In other words, these tools are positioned to provide a means to fine-tune litigation strategy, without necessarily affecting a sea change in how litigation gets done. Useful as these tools have so far proven, a major problem (to date) lies in the quality of the data used as input, which of course can compromise the accuracy of the outcome analysis. This has been particularly true in the US for those tools that rely heavily on data extracted from Pacer, which is notoriously inconsistent in the way it tags and classifies caselaw information. Garbage in/garbage out is still the operative principle by which human and artificial intelligence systems operate.
But the next generation of AI/ML software promises to dramatically advance the power and quality of these analytic tools. Notwithstanding the flaws in Pacer data, the next generation capabilities of natural language processing and AI algorithms will inevitably significantly improve the quality of data and reliability of the analytics tools now on the market. And it will take the idea of litigation outcome prediction to a whole new level. It will bring us to the point where a black box tool with oracle-like powers is capable of rendering the traditional approach to litigation more or less obsolete.
The broad outlines of this sort of outcome prediction capability are already in place. In fact, several litigation funding companies have already deployed AI engines to help them undertake the first pass analysis of pending lawsuits in order to identify those which are the best match for their portfolios. In the coming weeks, PinHawk will take a closer look at Apex Funding and Legalist, two of the litigation funders who are already successfully using AI/ML technology to help build their investment portfolios. But for now, just take a minute to think about what this means for the business of litigation in general – the fact that AI-powered software is now capable of reviewing a complaint and associated litigation materials and making a data-driven assessment of the likely case outcome. And moreover, this software has been tested and shown to be more accurate in predictive power compared to assessments made by highly experienced lawyers reviewing the same materials.
How can such software not have a dramatic impact on the way general counsel and boards of directors go about making decisions about which lawsuits should be brought and which vigorously defended? Of course, companies will still have strategic reasons to pursue litigation (such as delay and wearing down an adversary) even when a favorable outcome in court is deemed highly unlikely. But even so, it’s not hard to imagine a future in which it would be the equivalent of malpractice not to consult an AI-oracle before filing papers in court.
A recent article published on the Solomonic blog underscores how the reinvention of the business of litigation is already well underway. Solomonic is a UK start up that is on the leading edge of developing these AI tools with predictive power. One of their newest products, targeted for release later this month, will provide individuals and small businesses with an AI-powered online tool designed to help resolve employment disputes. In other words, it’s a black box tool that will potentially let parties cut to the chase with a highly automated form of dispute resolution.
As another recent Solomonic blog post suggests, with the advent of these analytic and outcome prediction tools, as AI-powered technology becomes capable of increasingly sophisticated legal analysis, solicitors should position themselves “more like management consultants” rather than traditional litigators who are always rushing off to do battle. In the words of Alex Oddy, a partner at Herbert Smith Freehills who is quoted in the article, these AI-tools will inevitably bring about “a change of mind set among lawyers” as they become more accustomed to the role of being “super forecasters.”
My only quibble with Alex Oddy is that I think he understates the significance of AI-technology when he describes it somewhat mildly as bringing about a change in mindset. It strikes me as the sort of change in mindset that accompanies a dramatic shift in the tectonic plates. True enough, even using the best seismic monitors, it remains impossible to know when an earthquake will happen. But we nonetheless can be pretty certain, sooner or later, it is coming.
Recent AI News