A new guide seeks to streamline AI’s role in e-discovery.
- The International Legal Technology Association (ILTA) released a draft guide on AI active learning in e-discovery.
- This guide aims to minimise conflicts arising from AI’s application in legal proceedings.
- ILTA’s guide awaits approval from the Civil Procedure Rules Committee and Master of the Rolls.
- Active Learning (AL) offers dynamic AI application, necessitating a clear and consistent process.
The International Legal Technology Association (ILTA) has introduced a draft best practice guide detailing the application of active learning (AL) in e-discovery processes. This technological advancement in artificial intelligence (AI) allows systems to learn from ongoing data analysis, thus modifying predictions as more documents are reviewed. Dr Victoria McCloud, a retired High Court master, emphasises that in the evolving AI legal landscape, especially given current court backlogs, this guide could significantly reduce disputes over active learning techniques.
As outlined, the guide is designed to encourage the adoption of AL technology by proposing a coherent process for legal professionals to follow. A spokesperson from ILTA highlighted that AL facilitates e-discovery by dynamically self-adjusting to document datasets relevant to each case. Consequently, it enhances the efficiency and effectiveness of legal document reviews. However, comprehensive understanding and transparency regarding its application are crucial, as AL determines the scope of disclosed documents.
Current procedural rules endorse e-discovery software, yet specific guidance on AI’s role within AL remains absent. The ILTA’s document was collaboratively developed over six months by litigation support experts from prominent law firms such as A&O Shearman, Linklaters, Norton Rose Fulbright, and DLA Piper. AL, identified as a process where initial document tagging aids in predicting the relevance of other documents, distinguishes itself through its capacity to continuously update predictions during reviews.
Though less common in contentious document reviews involving fewer than 20,000 documents, AL has demonstrated efficacy in significantly smaller reviews, provided validation processes are not mandated. Importantly, legal teams are advised to employ AL on document repositories exceeding 50,000 items, with consideration for pools above 20,000. Planning involves reviewing roughly 30% of documents within available timeframes.
Quality assurance and control form critical aspects of the guide, asserting that the quality of active learning is contingent upon initial teaching. The guide recommends meticulous oversight during the initial review days to standardise initial document assessments. New reviewers should be wary of over-classifying documents as relevant, with established workflows to address and rectify this bias if necessary.
The ILTA’s draft guide represents a pivotal step towards clarifying AI processes in legal contexts, promising to alleviate judicial burdens and enhance procedural consistency.
