Predictive coding has received a lot of attention lately as the next great magical wand in the e-discovery bag of tricks. However, as with any new technology, there are a number of different implementations and marketing claims that are confusing the whole picture of how this system can help make the e-discovery process more efficient and ultimately reduce costs.
In a nutshell, predictive coding involves the application of sophisticated artificial intelligence to permit the computer to make suggested determinations based on human interaction and the content of documents.
All predictive coding incarnations basically involve the review lawyer coding a subset of the records in the collection. The system examines the decisions made by the reviewer and identifies properties of the documents that it can use to automatically make determinations. As the reviewer continues to code documents, the system predicts what the reviewer will code. When the system’s predictions and the reviewer’s actually coding coincide (within reason), the system has learned enough to make confident predictions on its own.
Predictive coding is being applied at several stages in the e-discovery analysis and review processes:
Culling: In this mode, a lawyer who is an authority on the matter makes relevance decisions on a subset of the records. Once a sufficient number of records have been reviewed (typically a few thousand), the system applies its predictive analysis to the entire set to cull out the records most likely to be relevant. These records can then be subjected to the normal, manual review process.
Subjective Coding: The predictive coding system examines the subjective coding decisions made by lawyers as they manually review records. When a sufficient number of records have been reviewed, the system will start to make coding suggestions for subsequent records to assist the lawyers.
Review Quality Control: Along the same lines as predictive subjective coding, the system uses the subjective coding decisions made by lawyers to predict how documents should be coded. However, instead of suggesting codes for un-reviewed records, the system will apply the predictions to all manually coded records and identify those records where its predictions and the actually coding diverge. This will enable reviewers to zero in on documents that may not be coded correctly.
Prioritization of Records for Review: Predictive coding can also be used to prioritize records in a review. Once a sufficient number of records have been manually reviewed and coded, the system can group un-reviewed documents based on its coding predictions. The review project manager can then group all documents likely to be coded relevant, for instance, and assign these to be reviewed first.
Predictive coding technology is also being considered in several electronic records management solutions to permit automatic classification of records, removing the burden from individual users.
This technology is being incorporated into more and more e-Discovery software systems, and may soon become a standard way to cull and review electronic data.
For more information on this technology and other cutting-edge e-discovery solutions, contact us.