Inside Arbitration: Data analytics in international arbitration: Balancing technology with the human touch

Any form of dispute resolution can arguably be described as nothing more than the structured presentation of data analytics. Arbitration is no different.

At the outset of a claim, information (data) is collected and analysed to determine the facts of relevance to a dispute. Contractual and statutory documentation is reviewed to ascertain the applicable legal framework. Individuals are interviewed to capture information not recorded in contemporaneous documents. Legal counsel and arbitrators are selected and appointed based on the parties' (or an arbitral institution's) understanding of their experience, credentials, any conflicts and (where such information is available) their previously stated views on issues in dispute between the parties.

As a case progresses, vast swathes of documents are harvested, exchanged and reviewed. Relevant case law is analysed with a fine tooth comb, witnesses' testimony is presented and challenged and – ultimately – a legal case which draws together a (hopefully consistent) legal and factual story is presented in writing and, more often than not, at an oral hearing. In turn, the arbitral tribunal will process the data submitted to it and, based on its own analysis of that information, rule on the dispute.

Data analytics in Arbitration

Each of the above steps involves the analysis of data and over the course of an arbitration all four primary types of data analytics are required and used:

• Descriptive analytics help to answer questions about what happened, drawing on potentially large datasets to provide the essential insight into what has occurred in the past.
• Diagnostic analytics help to answer the question of why those things happened in the way they did. The techniques take the findings from descriptive analytics and dig deeper to find a cause. 
• Predictive analytics help to answer the question of what may happen in the future. These techniques draw on historical data to identify trends and determine whether they are likely to recur.
• Prescriptive analytics help to answer the question about what should be done. Drawing on predictive analytics, decisions can be made about the best way in which to proceed (e.g. how best to present an argument or case to a particular tribunal).

Historically, the analysis of data in an arbitration was carried out manually (by humans), drawing on their individual skill and expertise and those of their colleagues.

Lawyers would manually review all data received from their clients, laboriously research narrow points of law, study relevant publications or conference notes from tribunal members on salient issues (or rely on 'word of mouth' based on colleagues' past experiences). However, going forward technology will play an increasingly central role in these processes and will help to make the relevant analysis more efficiently and effective.

This will enable lawyers to take decisions in a more data-driven way, offer greater certainty to their clients (e.g. in relation to cost estimates or prospects of success), foster early settlement between parties in dispute, select more efficient or suitable decision-makers and better guarantee that an award can ultimately be enforced, if needed.

The transition: using technology to making human processes more efficient

In recent years, LegalTech across the entire legal sector has boomed. Within this, a large number of products have been designed to help arbitration practitioners (and litigators) streamline existing (largely human) processes. Those products include software tools that help amplify human review capabilities in the face of growing volumes of unstructured data and tight deadlines, help predict the behaviour of courts, judges/arbitrators, lawyers and other arbitral participants like expert witnesses, streamline legal research, produce first drafts of standardised documents, suggest indexes for hearing bundles or help estimate the likely length, cost and complexity of a given arbitration.

These legal analytics tools rely on technologies like machine learning and natural language processing to clean up, structure, and analyse raw data to identify and interpret patterns and trends within the relevant dataset. While the output that they achieve may be akin to the result of a human-led process, the means by which they achieve it is usually very different.

New tools have been designed to assist with each of the 4 types of data analytics explained above. However, many of those tools remain 'point solutions' aimed at improving a specific (and often narrow) process within the arbitration. A period of consolidation within the LegalTech market can be expected in years to come, mirroring what has happened in other sectors. Similarly, platforms and software tools will become more intelligent and better able to present information, as datasets increasingly become more structured and integrated as well.

Even now, the use of technology in the arbitration process is no longer a luxury and arbitration practitioners (including arbitrators, in particular, who are called upon to make decisions on what technology should be implemented and how) need to understand the capabilities and limitations of these tools. They also need to grasp the new, often complex, legal and regulatory issues that arise from the need to analyse increasing volumes of data relevant to and generated in an arbitration.

Indeed, insofar as technical upskilling is concerned, arbitration practitioners need to have sufficient digital literacy and data science skills to understand what datasets are being interrogated and how outputs are produced. If arbitration participants fail to grapple – at a high level at least – with the functionality of the tools used in the arbitration, this may lead to unanticipated or undesirable results (due, for example, to potentially unforeseen biases in the algorithm through which relevant trends in the datasets are identified). Data analytics tools often tend to focus on data that is most readily available (e.g. older awards and judgments that are publicly available). In the context of arbitration, which is a private and largely confidential process, available data may not therefore give the full picture.

As to the new legal and regulatory issues associated with the use of electronic data analytics tools through the arbitral process, relevant data protection laws need to be identified and complied with insofar as any personal data is being processed. Similarly, adequate steps must be taken to secure the data exchanged in an arbitration and ensure, both legally and in practice, that issues of control, possession and responsibility for the cybersecurity of the data are adequately addressed and that responsibility can be properly allocated if an arbitral participant falls short of the required standard of diligence.

The road ahead: moving from replication of human processes to a more disruptive technology-driven redesign of arbitral processes

The capabilities of data analytics software will continue to grow exponentially over the years to come. As software develops and becomes more sophisticated, the accuracy of its descriptive, diagnostic, predictive or prescriptive analytics will improve significantly. The adoption and use of these tools will become more widespread in arbitration, particularly for large complex disputes involving significant volumes of data and justifying the associated costs of licensing these tools.

As the accuracy of these tools gets better, parties will increasingly be driven to identify and settle disputes at an early stage. The tools will enable a more detailed calculation of return of investment on the time and cost involved in proceeding to a trial. However, new opportunities arising from emerging technologies and new applications of existing technologies may lead to a more fundamental redesign of arbitration processes.

Indeed, with arbitration being a creature of contact, this dispute resolution mechanism is a more likely to see party-driven change in the short to medium term than court based litigation for instance. The end users of arbitration are already driving greater efficiency in the arbitral process thanks to tools already on the market. In the years ahead, through the bundling of machine learning, natural language processing, internet of things (IoT) and other 'smart' devices and the huge processing capabilities of cloud and edge computing and, looking ahead, quantum computing, we will no doubt see some radical shifts in the processes proposed for resolving disputes through arbitration.

Might human arbitrators be supported by new software or hardware in their decision making, as is the case already for surgeons both at the diagnostic and surgery stage? Will a legal-led approach to dispute resolution be displaced by a 'game theory' analysis of party's positions? Or will human discretion and creativity remain a necessary part of dispute resolution to ensure that the law can continue to evolve and that fairness is achieved in all cases (including those outlier cases which do not follow a trend borne out in historical data)?

Lawyers need to engage with these questions today and work hand in hand with technologists and data scientists to help ensure that the dispute resolution methods of the future are fit for purpose and continue to have legal weight so that any ultimate decision or direction is valid and enforceable in the 'real', physical world.

Conclusion

Data analytics are nothing new in the context of arbitration. However, technology is already transforming the way in which those analytics are performed and making existing processes more efficient and effective. Going forward, new technologies or new applications of existing technologies stand to disrupt the arbitral process more significantly.

Some lawyers may resist this transformation out of fear that new technologies might change how they practice law or even make their jobs obsolete. Similar concerns were voiced when legal research moved from books to computers. However, that transition did not reduce the need for lawyers skilled in legal research and analytical reasoning. Instead, it enabled lawyers to be better and more effective at their jobs.

Similarly, technology-driven data analytics will not make the judgment and expertise of experienced lawyers obsolete. It will, however, enable those who employ such software to provide better and more cost-effective representation for their clients and better tailor their advocacy to the relevant audience. The widespread adoption of these new and emerging technologies in businesses across all industries will also generate new types of disputes which will continue to keep lawyers (and their artificially intelligent assistants) busy for years to come.