FORUM: Impact of AI and technology on litigation
October 2017 | SPECIAL REPORT: INTERNATIONAL DISPUTE RESOLUTION
Financier Worldwide Magazine
October 2017 Issue
FW moderates a discussion on the impact of AI and technology on litigation between Tony Sykes at IT Group UK Ltd and Darren Pauling at KPMG.
FW: To what extent are artificial intelligence (AI) and technology making their presence felt in the courtroom? How do these factors influence the way litigation proceedings are conducted?
Sykes: The most obvious use of AI in the courtroom is the increasing and significant use of software platforms that enable electronic trials and arbitrations to take place. The proliferation of offerings in this sector has meant truly paperless trials and real cost savings. The ‘intelligence’ of these systems often surfaces much earlier than the trial itself. It is in the review arena that AI has had the greatest impact. When documents have been required to be scanned and then passed through an optical character recognition process (OCR) or have been translated from a foreign language or recovered from a deleted file, then errors in the exact results are not uncommon. Algorithms have been developed and refined that enable searches to have a ‘fuzziness’ so that these errors or uncertainties do not result in key evidence being missed by the search engine. The review process itself, now considered to be one of the single most costly aspects of disclosure, can be automated by the use of AI. A sample of documents is reviewed by a review panel and then the AI algorithms ‘learn’ which type of documents trigger a review criterion and then the AI software completes the rest of the review.
Pauling: It sometimes feels as though ‘AI’ is a special label given to some magical technology which is just out of reach, and once we have it, it is no longer ‘AI’. The reality is that there is a spectrum of AI-related technologies which have been in use for 10 years or more. It is the role of a good technology partner to advise on which are best suited to a particular use-case, and to ensure they are applied in the most effective manner. We see a broadening acceptance of AI technologies beyond traditional use-cases. In competition or M&A for example, near duplicate detection and predictive coding technology can be applied to identify, sort and remove intellectual property which was to be retained by a different entity. This marks a shift from providing the ‘raw material’ for litigations to using technology to put into practice the legal solution required by the court.
FW: How would you characterise the pros and cons of introducing a greater degree of automation into litigation proceedings? Could you provide specific examples as to when this has aided or, conversely, hindered the process?
Pauling: The primary motivation remains cost reduction. Automation is a game changer, with ‘keys to the warehouse’ disclosures, where guided, live explorations using visual analytics allow SMEs to address key issues quickly and accurately. In one recent matter a party was presented with nine terabytes of email and electronic documents with no background or relevance to the matter. Through the use of conceptual analysis driven by SMEs, the data was quickly sifted and key material found that enabled a substantial litigation to settle early. Automation by predictive coding is a way to introduce a greater degree of consistency in the classification of documents. Applying this methodology and comparing a manual approach, the accuracy of reviewers can decrease as the review continues, attention spans shorten and minds wander. It remains imperative to ask the right questions, challenge assumptions and warn of potential pitfalls of any project.
Sykes: The benefits of automation are obvious in some aspects of litigation. Good e-disclosure software can enable evidence from huge email accounts to be found and presented with significantly less review time and with much improved confidentiality. Online processes for small claims have provided access to justice for many small traders and consumers who otherwise may have just given up on a dispute with a larger commercial organisation. There is a great divide, however, between automation that is available to litigants in person or very small law firms and the platforms run routinely by the large practices for most of their cases. The latter, while presenting real opportunities for saving, mostly require significant investment and training. The exceptions are web-based review platforms that require far less investment in training and can be deployed to a wider range of cases.
FW: To what extent can technology effectively assist in the collection of evidence, coping with large volumes of data from multiple sources, and controlling related costs? Furthermore, what factors determine the technologies available to litigators and how these are utilised in the courtroom?
Sykes: Traditional evidence collection for the purposes of disclosure has required litigants to perform their own trawl of the data in their possession after heeding the advice of solicitors as to what is and what is not disclosable. The huge growth in forensic IT that has happened in the last 30 years or so, in direct response to the use of computers and technology by criminals, has provided a range of skills and products that have enabled the harvesting of large quantities of data from many disparate sources easily and relatively cost effectively. The challenge today is how to present the key data from within that huge harvest to litigators in a form that engages with their processes and aligns with their skills and resources. With judicial services around the world striving to keep the civil justice system as cost effective as possible without jeopardising access to justice, the low-hanging fruit that is basic e-disclosure and the paperless trial have become go-to considerations. The dangers are that uncontrolled e-disclosure can result in tens of millions of documents being presented and costs simply shift from bundle creation to review.
Pauling: Data sets are ever expanding and to combat this growth sophisticated culling and identification methodologies can be deployed early in the process to help limit costs. Technology can be applied both before and after the collection itself. One recent example involved reducing the number of in-scope custodians from 60,000 down to 8000 by novel application of social interactions analysis and tracking of data dissemination through the business. Post-collection, text analytics and other early case assessment tools can give litigators a head start, going beyond keywords to consider themes, concepts and clusters within the data. This process can also flag potential gaps in the evidence collected. An experienced technology partner will start not with the technology itself but from a thorough understanding of counsel’s thinking and priorities. This will drive the technical solutions best suited to achieving those goals in the most timely and cost-effective manner.
FW: Going forward, how do you see greater use of AI and technology shaping litigation? Are they set to alter the process in any fundamental ways?
Pauling: We increasingly find that keyword-based approaches break down as data volumes increase. An alternative is to let the data tell its own story using visual analytics. This exploration is best conducted live, bringing technologists, litigators and client stakeholders together to work with the raw material, getting right to the heart of the matter. Another area of focus is investment, both in technical infrastructure and in the advanced skills needed to extend the off-the-shelf capabilities of many e-discovery tools, which will become a differentiator among technology providers. We also see many potential economic and informational benefits in a portfolio-based approach to litigation technologies, as enabled by newer ‘managed services’ arrangements with larger organisations. These include the re-use of collected material across matters, and automated extraction of richer knowledge graphs from those larger corpuses. A recent example involved a data remediation exercise where the material was reused as part of a regulatory investigation.
Sykes: Despite the astonishing advances in technology that have driven the e-disclosure phenomenon, most people are completely unaware of the subject. The headline grabbers are the AI algorithms that are being trialled and, in some cases, actually used to decide on sentencing in criminal cases. There are few arenas where human decision making is more closely scrutinised than in sentencing and yet there are increasing reports of AI being used to aid or even to replace that human element at the most crucial of stages of the judicial system. Algorithms that predict or report on the likely behaviour of a defendant when considering bail or the appropriate type of custodial premises have been in use for some time, particularly in the US. One of the most well-known applications and the system at the centre of a high-profile challenge to a Supreme Court ruling is Compas, a suite of software modules. Compas has a number of AI algorithms for the prediction of risk. Its use in the case of Wisconsin v. Loomis created controversy because the output from the AI algorithm was used to decide on the length of custodial sentence and yet neither the prosecution nor the defence had access to the algorithm. As Big Data and the algorithms that act on it become more established, a greater use of automated risk assessment and even sentencing is possible in other jurisdictions around the world.
FW: In your opinion, how far can the adoption of AI and technology in the courtroom conceivably extend? Can we seriously be looking at litigation and the pursuit of justice as a predominantly digital endeavour? How real is the possibility that litigation proceedings may eventually become too reliant on technology, such as using AI ‘judges’?
Sykes: While the Loomis case has raised many questions, there is undoubtedly a concerted move toward the adoption of risk assessment tools in many states in the US. While at present, paperless trials are on the increase in the UK, it can be argued that this is not AI at work but just a general computerisation or the relentless march of technology. However, the next stage in paperless trials is possibly to embrace AI so that the progress of a trial or arbitral hearing is controlled or adjusted, dependent on the way the evidence unfolds. In our experience, large arbitrations and long civil court hearings invariably involve many late night discussions between counsel and lawyers and their clients to decide on whether to call certain witnesses or which evidence to go to given the way the previous day has gone. The day that AI can assist in determining the strategy and therefore the predicted outcome of the hearing is not far away. While an AI judge may ultimately become another form of ADR, binding or otherwise, it is hard to believe that the judiciary generally will surrender its supreme position to a suite of algorithms and the ever-increasing mass that is Big Data.
Pauling: Technologists still drive the technology using the ‘satnav’ of the legal profession for guidance and direction. The stories told by data will always be for a human to interpret, and as analytical methods become more complex, the expertise needed to do so becomes even more important. Certainly there is no existing technology which could come close to exercising ‘judgment’. The current generation are really only advanced ‘recommender’ systems, applying previous human decisions to similar items: at best a simple form of ‘narrow’ or ‘vertical’ AI, designed to perform well within one context. The vision of AI with whom one could hold a ‘normal’ wide-ranging conversation, the ‘general AI’, is still out of reach. Even should they succeed however, dispensing justice often requires more than knowledge of the law; it is a long way from AI to artificial wisdom.
FW: What overall advice would you offer to parties in terms of managing the issues raised by the use of AI and technology in the courtroom?
Pauling: It should be acknowledged that these concepts and possibilities are not exotic, ‘the future is already here, it is just not evenly distributed’. The benefits of advanced processing workflows, analytics and visualisations are available now, but in many cases, are not being leveraged. It cannot be overemphasised that there is no ‘magic button’ which will select and disclose material without oversight. Everything is guided by the legal strategy, and there is ample scope for quality control, just as with traditionally-prepared disclosures. One can start incrementally, getting comfortable with descriptive analytics and exploring cluster visualisations, before trialling more advanced predictive usages. We have seen many cases where scepticism of the technology has been overcome by careful adoption of a workflow on a subset of the data backed up by appropriate human verification. It has been our experience that the machine is less likely to make a mistake if trained correctly.
Sykes: At present, parties presented with an option to use technology in the courtroom should consider cost first and then any other benefits as secondary considerations. The use of e-disclosure in the proceedings up to the trial itself is well established and tried and tested. The use of technology for paperless trials is less well advanced but in our experience some systems are extremely good but relatively expensive to operate and only find true benefits in larger trials. A 2016 study by University College London, the University of Sheffield and the University of Pennsylvania looked at the judicial decisions of the European Court of Human Rights and using AI they had developed, they predicted the outcomes of the cases to an accuracy of 79 percent. The leader of the study though did not see AI replacing judges or lawyers but the feeling was that AI could be useful for identifying patterns in cases that lead to certain outcomes and for highlighting which cases were most likely to be violations of the European Convention on Human Rights.
FW: Given the continued adoption of AI and technology in litigation proceedings, what steps should parties take at the outset of a dispute – or even before one surfaces – to incorporate technology into their preparations?
Sykes: When a dispute looms or is anticipated in some form, the preservation of data is key. While most larger organisations routinely store email and project documentation for agreed minimum terms – the creation of a vault of data for projects in particular is sound advice. We have seen many technology projects that have developed into disputes becoming extremely protracted and even frustrated because the platform upon which all the data, including email correspondence, was in the control of one party and inaccessible to the other once the project broke down and became the subject of proceedings. Organisations would be in a better position to handle disputes if and when they arise if consideration has been given in advance to the levels of technology that will be considered when a dispute arises. Simple steps to preserve machine readable versions of documents removes the need for expensive and time-consuming scanning and ‘OCRing’ and project-based backup of data, emails and configurations makes the identification of relevant data much more straightforward if a dispute develops sometimes years later.
Pauling: Early consultation with technology partners and with each other, on goals, feasibility and expectations will save headaches down the line. It can also help to include clients and subject matter experts in these discussions, as their familiarity with the data can help technology partners determine which tools are likely to provide the best insights. Even before this point, a mature information governance strategy is critical to being able to respond with agility to legal issues arising. This would include at minimum a data map, and accompanying data preservation strategy. AI technologies mean it will be increasingly difficult to hide needles in haystacks. Clients can manage risk and potential costs by making sure they have robust data retention policies adopted globally and managed correctly. A good technology partner will be able to advise on specific strategies for meeting these basic points and wider information governance strategies.
Tony Sykes is a chartered engineer and a chartered IT professional. He has over 35 years’ experience in IT fitness for purpose, software IPR, licensing and outsourcing disputes. He is a highly accredited expert witness with experience of giving evidence in the High Court and international tribunals and arbitrations. He has been instructed by many leading firms in the UK, EMEA and the Far East. He can be contacted on +44 (0)20 7096 3791 or by email: tony.sykes@itgroup-uk.com.
Darren Pauling is a managing director in KPMG’s forensic technology investigation practice in the UK. He has in excess of 20 years’ investigative experience and has overseen the development of KPMG’s forensic technology practice over the past 12 years. He has led multiple global litigation matters, investigations and more recently global remediation cases in relation to data separation and IP remediation. He can be contacted on +44 (0)20 7694 5565 or by email: darren.pauling@kpmg.co.uk.
© Financier Worldwide
International dispute resolution
FORUM: Impact of AI and technology on litigation
When to provide for litigation, arbitration, mediation or a combination of each
Overview of third-party funding internationally
Is international arbitration a reliable forum for dispute resolution?
International enforcement of judgments and arbitration awards
Witness evidence in international commercial arbitration
A note to general counsel: protecting foreign investments using investment treaties
The 2016 Russian guidelines on future BITs: have your cake and eat it