• 

Use Of Predictive Coding In Document Review For Arbitration

📌 1. What Predictive Coding Is in Arbitration

Predictive coding refers to the use of artificial intelligence / machine‑learning tools to assist in document review. A human reviewer first codes (“tags”) a sample set of documents as relevant/not relevant. A machine learning algorithm then learns from that sample and scores the rest of the document universe for relevance, vastly reducing manual review time and cost. It’s sometimes called Technology Assisted Review (TAR) or Computer Assisted Review (CAR).

In arbitration, predictive coding can be deployed in:

Pre‑arbitration case assessment

Document production exercises

Responding to tribunal orders for e‑disclosure

Privilege/redaction review

Privilege and responsiveness quality control

Because many arbitrations involve large volumes of electronically stored information (ESI), predictive coding can make disclosure proportionate, efficient, and defensible when properly implemented and agreed with the tribunal and other parties.

📌 2. Foundational Case Law on Predictive Coding (Litigation)

Although arbitration‑specific “predictive coding decisions” are rare, courts in several common law jurisdictions have accepted its use — and tribunals in international arbitration often look to these authorities on technology‑assisted review protocols.

(1) Pyrrho Investments Ltd v MWB Property Ltd [2016] EWHC 256 (Ch)

Jurisdiction: England & Wales High Court (Chancery Division)

Significance: First English case to expressly approve predictive coding in e‑disclosure.

Facts involved millions of electronic documents, and the parties sought court approval for a predictive coding protocol.

The court permitted predictive coding where the parties agreed on the methodology and it was proportionate and cost‑effective.

This decision is frequently cited as the landmark predictive coding case in the UK context.

Key Takeaway: Even without arbitration rules explicitly addressing TAR, courts will endorse predictive coding where it promotes proportionate and efficient document review.

(2) Brown v BCA Trading Ltd [2016] EWHC 1464 (Ch)

Jurisdiction: England & Wales

Significance: First reported English case ordering predictive coding despite one party’s objection.

The High Court ordered the use of predictive coding for disclosure, emphasizing cost considerations and proportionality relative to keyword / manual review.

Key Takeaway: Courts can mandate predictive coding even over objections where it promotes fairness and efficiency.

(3) Da Silva Moore v Publicis Groupe & MSL Group (S.D.N.Y., 2012)

Jurisdiction: United States District Court (SDNY)

Significance: One of the earliest endorsements by a U.S. federal judge of predictive coding in e‑discovery.

The opinion approved a predictive coding protocol for a large document review exercise and addressed proportionality concerns.

Key Takeaway: U.S. courts recognize predictive coding as at least equally defensible to other review methods, especially where ESI volume is high.

(4) Irish Bank Resolution Corporation Ltd v Quinn [2015] IEHC 175

Jurisdiction: Ireland (High Court)

Significance: Irish High Court endorsed the use of predictive coding where parties agree to its use.

This decision was cited with approval in Pyrrho as an international authority on TAR.

Key Takeaway: Predictive coding acceptance spans multiple common law jurisdictions.

(5) McConnell Dowell Constructors (Aust) Pty Ltd v Santam Ltd & Ors

Jurisdiction: Supreme Court of Victoria (Australia)

Significance: Court endorsed a predictive coding agreement and special referee review protocol due to the huge volume of discovery documents, and noted that predictive coding is efficient and proportionate.

Key Takeaway: Australian courts likewise validate predictive coding for large ESI inventories, supporting broader use.

📌 3. How Predictive Coding Applies in Arbitration

Unlike litigation, arbitral rules (e.g., ICC, LCIA, SIAC) generally do not yet contain express provisions on predictive coding. However:

Tribunal case‑management powers implicitly allow parties to propose technology solutions, including predictive coding, where appropriate.

The IBA Rules on the Taking of Evidence in International Arbitration (2010) encourage a party to use proportionate and efficient search methods for document production — a principle embodied in predictive coding usage.

The ICC Commission Report on Leveraging Technology acknowledges predictive coding as a legitimate tool for identifying responsive documents and advises that its use, disclosure, and protocols should be discussed at case‑management conferences.

Arbitral tribunals will generally consider predictive coding in light of:

✔ Proportionality of costs
✔ Volume and complexity of ESI
✔ Agreement between parties or need for tribunal approval
✔ Transparency of algorithmic process and quality‑control protocols
✔ Non‑discrimination between parties in access and review quality

📌 4. Why Predictive Coding Matters in Arbitration

Below are core reasons predictive coding is increasingly relevant in arbitration document review:

🔹 Efficiency and Cost‑Effectiveness

In disputes with millions of documents, manual review is costly and slow. Predictive coding can dramatically reduce human review time and expenses.

🔹 Improved Accuracy and Consistency

Statistical evidence shows predictive coding often matches or exceeds the relevance retrieval of keyword searches and manual review.

🔹 Proportionality

Tribunals must balance discovery burdens against arbitral efficiency. Predictive coding supports proportional document review where appropriate.

🔹 Transparency and Defensibility

Parties should agree on predictive coding protocols (seed set size, quality control, stopping criteria) to ensure defensibility and avoid disputes over methodology.

📌 5. Practical Protocols for Predictive Coding in Arbitration

Although arbitration lacks uniform TAR rules, tribunals and counsel generally adopt predictive coding protocols similar to those used in litigation:

Early case‑management conference to discuss potential use

Agreement on seed set selection and training rounds

Specification of quality‑control measures (e.g., random sampling of non‑responsive documents)

Transparent reporting to opposing parties/tribunal

Documentation of methodology for defensibility

These steps mirror best practice from litigation predictive coding cases and guidance reports.

📌 6. Limitations and Challenges

Even with supportive authorities:

⚠ Tribunal approval may be required if predictive coding significantly affects production obligations.
⚠ Transparency might be contested if one side fears tactical disadvantage.
⚠ Quality control disagreements can lead to satellite disputes.

The case law above underscores that agreement or reasoned tribunal direction is key for predictive coding’s defensibility.

📌 Summary of Case Law Authority

📌 Conclusion

Predictive coding is not yet mandated by most arbitration rules, but its use is fully supported by major jurisprudence in litigation and by international arbitration guidance. Parties and tribunals should address its use early, agree on transparent protocols, and justify its proportionality. High‑profile cases such as Pyrrho and Brown v BCA Trading provide strong authorities showing that technology‑assisted review is acceptable, efficient, and defensible — principles that arbitral tribunals are increasingly willing to adopt.

RELATED Blog

Maritime Arbitration in India

Legal recognition of ADR in India

Arbitration law in Afghanistan

Arbitration Law in Albania

Arbitration Law in Algeria

Arbitration Law in American Samoa (US)

Arbitration Law in Andorra

Arbitration Law in Angola

Arbitration Law in Anguilla

Arbitration Law in Antigua and Barbuda