Adoption of Technology Assisted Review to increase worldwide following a decision by the Irish High Court

Introduction

In the Irish High Court case Irish Bank Resolution Corporation Ltd & Ors v Quinn & Ors [2015] IEHC 175, Justice Fullam has delivered a landmark decision paving the way for Technology Assisted Review ('TAR').

TAR or predictive coding is an alternative to the traditional linear1 review of documents and involves:

  • The linear review by senior members of a legal team of a sample 'seed' set of documents to identify whether they belong to certain categories (relevant, not relevant, privileged, etc.).
  • Computer analysis to apply the characteristics of the sample 'seed' set to the full population of documents to group them into the same categories. The resulting reviewed document set is consistently categorised using a process which is both auditable and repeatable.

For a more detailed discussion on TAR please refer to our previous Forensic Matters publications No.12-03 Is Predictive Coding the electronic discovery 'Magic Bullet'? (July 2012), and No.13-02 Death, taxes and computer assisted review (February 2013). Also note that Computer Assisted Review ('CAR') and Technology Assisted Review ('TAR') are interchangeable terms.

This Irish decision appears to be the first endorsement of TAR by a Common Law Court, and is further evidence of increasing judicial acceptance of predictive coding in jurisdictions across the globe. (See, for example, in the US, Da Silva Moore v Publicis Group, et. al.).

The decision is in line with the Sedona principles2, which in essence state that individual parties are the best people to evaluate the best process for producing electronic documents.

Background to the case

The principal plaintiff in this case, the Irish Bank Resolution Corporation Ltd ('IBRC'), is successor to the Anglo Irish Bank ('AIB') which is in liquidation. The defendants include Sean Quinn and his family (Sean Quinn was at one stage the richest Irishman in history). The plaintiffs claim that the defendants conspired to wrongfully convert bank assets of €455 million by arranging for AIB loans to Sean Quinn's wife, their five children, and ten other clients, for the purpose of buying shares in AIB in July 2008.

AIB hoped that these loans would help to unwind the significant stake Quinn had personally built-up (Quinn had discreetly purchased nearly 30% of AIB using complex financial derivatives), prop up the share price, and alleviate investor concerns about the viability of the bank. This occurred whilst the property market in Ireland crumbled, and borrowers were struggling to repay loans that AIB had made during the boom years.

Dispute

In this dispute the plaintiffs were seeking Court approval to use TAR in order to complete discovery of electronic documents. Their argument focused on TAR saving time, and being more cost-effective compared to the traditional linear review method of discovering electronic documents.

For the Australian context this is particularly interesting because we sense that lawyers here are held back from fully adopting TAR in the absence of Court validation of the process. KordaMentha Forensic has undertaken a number of engagements using TAR, where both the timeframes and the document numbers (in excess of one million items) have been challenging. We are not aware of any objections from opposing parties when our clients have implemented TAR.

The defendant objected to the plaintiffs using TAR on a number of grounds. Primarily that it did not comply with O.31 r.12 of the RSC3, but furthermore, the use of TAR raised other objections listed below in italics with our comments.

[47.1] "TAR will not capture all relevant documents and therefore is not compatible with the obligations of a party making discovery, which is the objective target of 100% of relevant documents".

The objective of TAR is to find as nearly as possible all relevant documents specific to a case and/or category. Thus the expectation is that predictive coding will be more accurate than linear review, and this has been established through studies4. The reality is that achieving 100% of relevant documents is an unrealistic objective – whatever method is used.

[47.2] "TAR is not suitable for data sets of less than 1 million documents".

In our experience, TAR is effective with all data sets, regardless of size. The main drivers are how well the client and law firm understand the documents, along with the experience of the review team that is creating the training sets. In this matter it was reported "that TAR is wholly appropriate and suitable for data sets of less than 100,000 documents. Dr. Khadvezic [plaintiffs' expert] says that the accuracy of the review is not compromised by using TAR for a smaller data set than 1 million documents".

[47.3] "The Court has not been told what the f-measure is going to be".

The f-measure is a user-determined ratio: a function of the risk of missing a responsive document offset against the time and cost to review documents. We suggest that the information should be disclosed as part of validating the process. In this case the plaintiffs did not disclose a specific f-measure but instead formed a view that the minimum f-measure should be 80%.

[47.4] "The Training Sets are not specific to the categories of discovery, and the sets might not contain any relevant documents".

Workflows are generally tailored to each dispute. We find that it is beneficial for the client to tailor a workflow around specific categories. However, the process outlined below by the plaintiffs may be just as effective, and is likely to have been tailored to suit the dispute in question.

"The plaintiffs propose a broad question namely 'is this document relevant to any discovery category?' The system will learn what is and what is not relevant. Dr. Khadvezic says the defendants' objection is unfounded because training the computer is a supervised approach. The first training set is created judgmentally, but subsequent training sets involve selecting documents which are most difficult to predict rather than being chosen at random. Furthermore, Mr. Crowley [plaintiffs' expert] states that attempting to intentionally teach the computer wrongly will not work. He cites a study – The Impact of Incorrect Training Sets and Rolling Collections on Technology Assisted-Review – which concluded that 'the impact of wrong training documents was smaller than expected; inserting up to 25% wrong documents resulted in only 3-5% less classification quality'. Furthermore, the control set is created by the system and has a 95% confidence measure."

[47.5] "There will be no savings in cost and time".

Even in undertaking TAR, manual review is still pivotal in determining relevance across the data set. However, you would still only expect to review a very small fraction of the case corpus to determine what the relevant documents are (potentially as little as 1%). By contrast, undertaking a traditional linear review would require each of the one million documents to be looked at.

In our experience, using TAR does not replace the requirement for senior legal persons to review documents. However, it does allow for prioritisation of documents by relevance, resulting in a far quicker and more cost-effective solution for the client. This is achieved by significantly minimising the review of non-relevant documents.

In this case, for 680,000 documents, the plaintiffs' legal representative "estimated that a traditional linear review, using a team of 10 experienced reviewers, would take 9 months at a cost of €2m leaving supervision and technology costs aside, whereas, the use of predictive coding would enable the plaintiffs to make discovery within a much shorter timeframe and at substantially lower cost".

Based on the numbers detailed by the plaintiffs, a linear review across this number of documents equates to a cost of €2.94 ($4.30 AUD) per document reviewed. TAR would save over €1.8million in review costs (assuming the plaintiffs' estimate that fewer than 10% of documents would then require review).

The case also gives some guidance on a number of practical areas when dealing with TAR. This includes the methodology and approach used by the plaintiffs, together with detail on dealing with commercially sensitive and privileged documents.

Conclusion

As is to be expected with any new concept, there has been healthy scepticism in adopting TAR, except in exceptional circumstances. We believe that this judgement provides the validation for predictive coding to be adopted in Common Law countries, including Australia. Given the cost and time savings that can be achieved it will be surprising if corporates and law firms do not adopt this technology.

At KordaMentha Forensic we have used TAR successfully to assist parties with their electronic discovery, in both small and very large cases. In all instances we have found that TAR provides a significant return on investment for our clients, providing positive outcomes from both a review and cost perspective.

It is interesting to note the plaintiffs in the Irish matter used the Clearwell Review Platform in conducting TAR. KordaMentha Forensic and its clients have been successfully using Clearwell and its Transparent Predictive Coding technology for a number of years.

The following is a link to this landmark judgement: http://www.bailii.org/ie/cases/IEHC/2015/H175.html

Footnotes

1A "linear review" is a traditional review of documents within an electronic discovery review platform. This means that the legal team, review team, or investigators, will look at one document after another, ordered by date or keyword relevance. This is very much a brute force method to reviewing the documents, looking at one document after another, until the entire data set is complete.
2The Sedona Conference Best Practices Commentary on the Use of Search and Information Retrieval Methods in E-Discovery, 8 Sedona Conference J. 189, 193 (June2007)
3Rules of the Superior Courts O.31, P.12 http://www.courts.ie/rules.nsf/0/cc15c27a9413d3a980256d2b0046b3db?OpenDocument
4Maura R. Grossman & Gordon V. Cormack, Technology-Assisted Review in E-Discovery Can Be More Effective and More Efficient Than Exhaustive Manual Review, XVII RICH. J.L. & TECH. 11 (2011), http://jolt.richmond.edu/v17i3/article11.pdf.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.