A key consideration when selecting AI technology for use in eDiscovery is to ensure that the purpose of the AI tool matches the desired results. All artificial intelligence tools are narrowly tailored to solve specific challenges. For example, most existing TAR technology is based on text classification and currently will not work well on documents that are strictly or primarily images (jpg, gif, etc). Many images with relevant content could be missed if TAR were to be applied. Steps should be taken to normalize the data to ensure it is properly prepared for the use of the selected AI tool. For example, optical character recognition (“OCR)” can be used to extract textual content from images. A consideration to bear in mind is: Garbage in, Garbage out. Without appropriate inputs, AI tools will not generate the desired results.
Another consideration when utilizing AI, as with all technology, is to ensure that sensitive data is secure during the entire process. A security team should be involved in auditing the entire AI data lifecycle as any unsecure links in the chain can compromise sensitive data.
Highly scalable solutions allow for increased potential use in legal analysis. They also allow for mistakes to rapidly scale across an entire data set. If the tool is not well-suited for the task, the data is not properly prepared, or there are too many coding inaccuracies in the human coding, such inaccuracies can promulgate across the universe. As a result, sufficient QC of human input is critical because AI has the power to amplify bad decisions, as well as good decisions.
This article has not been revised since publication.
This post was created by JenW on February 15, 2021.