AI and Construction Law: an essential and inevitable partnership
Part 2: Dispute resolution and predicting dispute outcomes
By Stacy Sinclair, Senior Associate, Fenwick Elliott
Wouldn’t it be great to have a crystal ball to see into the future and understand the likely outcome of your dispute, before embarking on that costly adjudication or court proceedings?
Whilst this off-the-shelf crystal ball is not yet available in stores for immediate purchase, some exciting developments have taken place in legal tech over the past few years. We are now starting to see the use of new technologies in dispute resolution and indeed new studies and research allow us to glimpse what might be just around the corner.
Part 1 of this series considered AI and construction law in the context of risk and contract management, and looked at a few of the technologies that are available now to assist in this respect. Part 2 now looks at the use of AI in the context of dispute resolution and predicting the outcome of disputes.
AI and Dispute Resolution
It is now commonly accepted that the industry often uses the term “AI” generally to cover discussions around machine learning, automation, pattern recognition within text and the automation of extracting this text. This is known specifically as “applied” artificial intelligence and is well used in applications that require the performance of a specific task and/or an automated, logic-based decision or action.
With regard to machine learning, this is a system or software which “learns” from the data it processes, through the use of algorithms. The software can learn from tags already applied to the documents (supervised learning) or it can categorise/cluster documents itself based on common characteristics (unsupervised learning). A system can also learn from the success of its previous decisions (reinforcement learning). In reinforcement learning there is no correct answer from the outset, but the system learns through trial and error when a user/reviewer says whether it is right or wrong, as it goes along.
“Strong” AI are those processes which are equivalent to human intelligence and have the ability to reason, make decisions and replicate human cognitive functions.
In the context of construction, whilst “strong” AI is perhaps some way off, “applied” AI certainly is here and is in use to some extent already. Machine learning technologies and AI-based data analytics are employed at various stages of construction and energy projects: contract formation, project management, manufacturing and construction and dispute resolution.
In terms of dispute resolution, to date, the disclosure process perhaps has seen the most visible benefits from machine learning.
Disclosure, the stage of formal litigation or arbitration proceedings where each party discloses to the other the documents that are relevant to the issues in dispute, requires the processing and review of potentially millions of documents, depending on the case. Historically, these documents would be manually reviewed by paralegals and lawyers – a lengthy and costly exercise. Now, with “predictive coding” (i.e. computer or technology assisted review), parties can employ machine learning technologies to train the software to assist with the review of the data set. Lawyers tag documents with a particular status (i.e. “relevant” or “not relevant”) and the software learns from this categorisation, identifying and tagging subsequent documents similarly. The software’s algorithm is constantly updating as it learns from either further tagged documents or corrections lawyers have made to the software’s previous output. Predictive coding allows for a more focused and efficient document review process.
At the moment, whilst the use of predictive coding is certainly increasing, there is still a somewhat hefty price tag for its use. Smaller, low value disputes are not necessarily able to justify this cost. However, with the rapid developments in technology these days, we may well see a change in this soon. Furthermore, the introduction of the Court’s new Disclosure Pilot may also increase the use in machine learning and AI-based technologies.
On 1 January 2019 a two-year Disclosure Pilot scheme commenced in the Business and Property Courts in England and Wales, which include the Technology and Construction Court (TCC). A new Practice Direction to the Civil Procedure Rules (CPR) applies and the aim of the scheme is to facilitate and influence a change in the approach to disclosure of documents in the litigation process – including a greater use of technology in the process. Parties are required to consider the use of analytics and technology or computer-assisted review tools as a means of expediting document reviews. Where they have decided against the use of such tools (particularly when the number of documents to review exceeds 50,000), parties must justify that decision.
Big Data and Analytics
Dispute resolution inevitably concerns the analysis of data. Lawyers need to understand the issues and evidence in the case, analyse the strengths and weaknesses of that case and advise on their client’s chances of success (amongst other things). Given the sheer amount of data generated each day (by next year the entire digital universe is expected to reach 44 zettabytes), the ability to analyse big data sets efficiently and effectively is of the utmost importance. Indeed, the ability to access, analyse and apply specific types of data could potentially have a strategic advantage to disputing parties.
Data sets of evidence in construction and energy disputes in particular can amount to many terabytes of data for each dispute – and no doubt with the advancement and increased use of new technologies in the design and construction of these projects, this is only set to rise.
Before turning to the resolution of disputes, in terms of dispute avoidance we are now starting to see AI-based, real-time analytics used on construction projects. For example, parties can jointly monitor and analyse metrics from on-site activities, allowing them to track and report transparently and instantly, and therefore react and adjust as needed. This may assist in avoiding or minimising the escalation of disputes.
In dispute resolution, the ability to analyse and harness big data efficiently and effectively may have a strategic advantage. The technology available now, including AI searching and clustering functionalities, can enable lawyers and their clients to interrogate big data sets and draw out patterns and connections in the documents and correspondence, and generally gain a deeper insight into the evidence and facts of the case. The technology currently being developed goes further and aims to provide parties with analytics to assist in the prediction of the outcome of the case. Data analytics and metrics which aid in predicting outcomes ultimately may shape the trajectory of a case, allowing parties each to decide whether to continue with the proceedings and/or at what point to reach a settlement.
Predicting the outcome of a case will depend on a number of both legal and non-legal factors. Having an understanding of these factors and access to data which analyses the influences of these factors on judicial decisions can be strategically advantageous: parties can make informed decisions during the course of the dispute process, manage expectations and possibly encourage early settlement.
Whilst the industry is only at the beginning of developing these technologies, there are platforms available now. For example, one UK solution has analysed the Commercial Court’s decisions and provides smart data and metrics on its judges and their decisions. The solution provides data on issues such as what percentage of a particular type of claim is likely to succeed. What is the success rate of s. 68 arbitration appeals? What is the success rate in real estate claims? With regard to the data on specific judges, for example, how has a judge ruled in the past on a particular issue and what is his or her willingness to disagree with previous decisions? The platform recognises that the identity of a particular judge may influence the outcome of a case and therefore success rates and other issues are also shown in relation to a specific judge. Another example is a US solution which provides analytics on California judges and their decisions.
In addition to emerging technology which provides metrics and smart data for informing decision-making during a dispute, there are also several recent studies which have sought to demonstrate the power of computers when it comes to predicting the outcome of disputes.
In October 2017 software developed by a Cambridge start-up company CaseCrunch predicted the outcomes of 775 PPI mis-selling claims. The software was asked to predict “yes or no” as to whether the financial ombudsman would succeed in the claim. The software had an accuracy of 86%. The 112 lawyers who analysed the same 775 claims had an average of 62.3%. CaseCrunch said that if the question is defined precisely, as was the case with the 775 PPI claims, “machines are able to compete with and sometimes outperform human lawyers”.
A further example is a study from researchers from University College London, University of Sheffield and University of Pennsylvania who were able to predict the results of human rights cases at the European Court of Human Rights (in respect of Articles 3, 6 and 8) with an accuracy of 79%.
Emerging AI-based platforms have the potential to transform the landscape of dispute resolution. Whilst we are only at the start of these exciting developments, it is clear that the use of analytics, big data and new digital technologies will enhance efficiency and efficacy in dispute resolution. The crystal ball is not yet available for purchase; however, solutions which provide smart data for lawyers and their clients to review evidence, make informed decisions and predict outcomes are rapidly evolving.