What is Predictive Coding?

Predictive coding is a general term for any technology which “predicts” which documents are relevant to disclosure issues based on training or instructions from a subject matter expert (usually the lawyer running the case).

The underlying technology behind predictive technology can vary: most predictive coding approaches were based on machine learning algorithms until 2024, when approaches powered by generative AI started to be introduced.

A key feature of predictive coding is validation: a human should validate the accuracy of the algorithm’s decisions so that the results can be defensible in court. Validation is based on statistics which measure false positives and false negatives in the algorithm’s decisions.

The use of predictive coding was approved in UK courts Pyrrho Investments Limited & Anr v MWB Property Limited and Others <2016> EWHC 256 (Ch).

Written by Rachel McAdams (Senior Consultant at Sky Discovery)