Technology — friend or foe?
In light of COVID-19, we are more reliant on technology than we have ever been in our everyday working lives. In this article we examine how practitioners can use Technology Assisted Review (TAR) to assist with the task of preparing discovery.
The discovery burden, particularly in large construction disputes, can be heavy in terms of time and cost because of the sheer number of documents that are produced during significant construction projects. As such, parties and lawyers are keen to explore how different forms of technology can be utilised to reduce costs. One of the options available to parties seeking to manage and review large data volumes is by using TAR. TAR — also known as Computer-assisted Review (CAR) or predictive coding — is an emerging trend in discovery across the globe. In this article, we explain what TAR involves and explore how TAR can be used within the parameters of the High Court Rules (HCRs).
The use of technology under the high court rules
The overall objective of the HCRs is to secure the just, speedy and cost-effective determination of a proceeding.
Consistent with this objective, the HCRs relating to discovery are designed to be flexible and responsive to the specific requirements and cost considerations of each case. In particular, sch 9, cl 3 of the HCRs contemplates the use of keyword searches, automated searches and techniques for culling documents (including concept searching, clustering technology, document prioritisation technology, email threading, and any other new tool or technique). This rule opens the gateway to parties exploring the use of TAR when completing discovery.
What is Technology Assisted Review?
TAR is an iterative process, whereby a human reviewer tags documents as relevant or irrelevant and the computer algorithm “learns” from that human input to identify other potentially relevant documents from the pool of unreviewed or unseen documents.
The older version of TAR (referred to as “TAR 1.0”) required a senior lawyer to review a sample set of documents to “train” the computer algorithm as to which documents are relevant and which are not. Once it has been “trained”, the algorithm undertakes an analysis of the entire document population and creates a list of documents that it considers to be “relevant”. Those relevant documents are then reviewed by a solicitor who is part of the team working on the case. The TAR 1.0 process was seen as a useful tool to undertake the first level review to determine which documents met the threshold of being relevant, following which the human reviewers could start their review from that new baseline.
The second and current version of TAR (referred to as “TAR 2.0”) uses a continuous active learning (CAL) algorithm whereby the algorithm continues to learn as the human legal team progresses with their review of the documents to ascertain whether the sets of documents are relevant/not relevant. CAL is more agile that TAR 1.0 because it regularly re-orders the document population based on what it has learnt — in practice this means that the algorithm is constantly promoting documents that it considers to be relevant to the top of the human reviewers’ list or set of documents. One of the problems with TAR 1.0 was that the one-time learning training method meant that all the documents had to be collected before starting the review — otherwise the inclusion of new documents rendered the original training invalid. The advantage with TAR 2.0 is that documents can be added on a rolling basis (that is, as and when documents are identified in a company’s records and provided to the solicitors acting in the matter) and be ranked by the algorithm according to their relevancy to the issues in the proceedings.
Parties may wish to engage external technical consultants before discovery is commenced to work alongside legal counsel and put together a framework and parameters around the use of TAR. Once parties wishing to use TAR have finalised a framework, then all parties should endeavour to agree upon the proposed frameworks for the use of TAR at an early stage of the case.
How does the use of TAR sit alongside the HCRS?
Rule 8.2(1) requires the parties to cooperate to ensure that the discovery process is facilitated by agreement on practical arrangements. Rule 8.2(2) further requires the parties to consider options to reduce the scope and burden of discovery and ensure that technology is used efficiently and effectively. The combination of r 8.2 and sch 9, cl 3 therefore provide express support for the use of technology in discovery and inspection, which would include the use of TAR.
In large construction litigation proceedings, where tailored discovery is preferred, the first step is for parties to try and agree upon the tailored discovery orders, failing which parties may seek formal orders from the court. In light of the requirement on the parties to cooperate in relation to discovery, the authorities suggest that the most logical time for the parties to agree upon discovery parameters, including how TAR would be best utilised, is at the time of seeking to agree upon tailored discovery orders. Failing agreement, the party or parties should highlight to the court that they wish to use TAR so that the court can refer to this agreement in a tailored discovery order. For example, in a leading English decision of the High Court in Pyrrho Investments Ltd v MWB Property Ltd  EWHC 256 (Ch) (Pyrrho Investments) the parties agreed upon the use of “predictive coding” in the disclosure process of three million documents but sought the Court’s approval to use the process, given its novelty in the English jurisdiction. The Court approved the use of predictive coding in the disclosure/discovery process on the basis that both parties agreed on its use and it was suitable given the significant number of documents that would need to be reviewed as part of the discovery process in that case. Helpfully, the Court outlined the key points that weighed in favour of it approving use of predictive coding, including that the parties had agreed to use the technology, the volume of documents that the parties would be required to review, the cost associated with a “manual review” (that is, a page-turning exercise by a human reviewer), the cost-saving measures that predictive coding could provide and, with reference to the timetable to trial, whether parties had time to consider other methods of disclosure if predictive coding turned out to be unsatisfactory.
The decision in Pyrrho Investments has set the scene for the use of TAR in cases where parties are unable to reach agreement on the subject. An illustration of this approach is visible in the case of Brown v BCA Trading Ltd  EWHC 1464 (Ch), where one party objected to the use of TAR, but the Court approved it by applying the factors set out in Pyrrho Investments. Of course, parties may not always be able to agree on whether TAR should be used, and importantly, how the technology should be used in a particular proceedings but in that instance, in accordance with the requirements in the HCRs, the parties should refer the matter to court for a clear determination (as would be the case with any other contentious discovery matter). In our view, the most opportune time for the matter to be referred to the court is at the time that tailored discovery orders are being sought by the parties.
Affidavit of documents: transparency around use of TAR
Rule 8.15 requires a party to file and serve an affidavit of documents. In particular, r 8.15(2)(c) requires parties to outline the steps that they have taken to satisfy the tailored discovery orders. Unsurprisingly, cooperation and agreement on appropriate parameters at the beginning of the discovery process can go a long way towards mitigating the risk of lengthy and costly interlocutory applications. In Irish Bank Resolution Corporation Ltd v Quinn  IEHC 175 at , the High Court in Ireland approved the use of TAR and stated that:
As technology assisted review combines man and machine, the process must contain appropriate checks and balances which render each stage capable of independent verification. A balance must be struck between the right of the party making discovery to determine the manner in which discovery is provided and participation by the requesting party in ensuring that the methodology chosen is transparent and reliable.
Failure to be transparent about the use of TAR, by seeking agreement from the other parties in the proceedings at an early stage, can minimise the need to make formal applications for want of compliance with the HCRs later down the track. This point was highlighted by the Technology and Construction Court in Triumph Controls UK Ltd v Primus International Holding Co  EWHC 176 (TCC). In Triumph, the parties agreed that certain keywords would be used across electronic document sets and that all responsive documents would be “manually reviewed” by a human reviewer. However, the plaintiff decided to use TAR in respect of undertaking its disclosure exercise without notifying the defendants. The Court held that the decision to use TAR was made unilaterally, without the defendant having any input into the process, which was unsatisfactory. The issue was compounded by the lack of information provided by the plaintiff as to the sampling exercise that was used and how this exercise was conducted by using TAR. On this basis, the Court made an adverse finding against the plaintiff, holding that the discovery process undertaken by the plaintiff was neither transparent, nor could it said to be independently verifiable. Ultimately, the Court ordered the parties to agree upon a methodology by which a sample of documents (in that case 25 per cent of the 220,000 retrieved by running search terms) was to be manually reviewed. The case demonstrates the consequences of obscuring the discovery process, and the adverse implications that could arise from a failure to cooperate when using TAR as a tool in the discovery process.
Utilising technology, including TAR, correctly when undertaking a discovery exercise has the potential to offer cost savings. This is particularly true for large scale construction disputes where the burden of discovery can be significant due to the volume of documents involved. Parties that find themselves in the position of having to review an enormous amount of data for discovery should explore how the use of TAR can create time and cost benefits. However, parties in New Zealand should bear in mind the leading decisions in England and Ireland, and the tenor of the HCRs which require parties to cooperate and engage in a transparent process that secures the just, speedy and inexpensive completion of discovery.
Christine Gordon is a Senior Associate in MinterEllisonRuddWatts' Construction team and Shukti M Sharma is a Senior Solicitor.
This article first appeared in the May 2020 edition of the New Zealand Law Journal.