New Study Shows AI Made Scientific Discoveries Humans Missed: What are the Implications for eDiscovery?
People are always talking about how Artificial Intelligence is the future of legal technology. But up to this point, very few are using it within eDiscovery, and those who are stick with the standard TAR / Predictive Coding approach. Outside of legal, breakthroughs are happening which could open doors for attorneys and investigators to use AI to help sift through large datasets while making connections otherwise not possible with human-only review.
In a recent study published in Nature, researchers from the Lawrence Berkeley National Laboratory used an algorithm called Word2Vec to read scientific papers. The algorithm was given no training in scientific knowledge, instead relying only on word associations. While reviewing over 3 million previously written scientific papers and looking for associations humans may have previously missed, the AI led researchers to knowledge that existed but wasn’t apparent without the help of machine learning.
Vahe Tshitoyan, the lead author on the study, stated, “This algorithm is unsupervised, and it builds its own connections.” Because it’s not trained on a specific dataset, you could easily apply it to other disciplines. He continues, “The information is out there. We just haven’t made these connections yet, because you can’t read every article.”
But any shadow of a doubt can throw a court case into question, which is why bringing AI into the legal sphere is tricky at best. This very well-considered and in-depth article in Law360 discusses why AI Tools need to be litigation ready or “discovery in a lawsuit contesting decisions those tools have made could quickly become a nightmare: Your company may suffer enormous distractions and decreased productivity as it struggles to address litigation requirements that are inconsistent with its AI systems, data and culture; may be subjected to onerous court orders that interfere with its ability to conduct its core businesses; may even suffer adverse judgments on claims that lack merit.”
This highlights the push-pull that exists around AI in the eDiscovery industry: innovators will create technology and find potential uses for it in legal, but the need for data integrity and defensible processes will slow those advances’ practical application.
But forward movement is forward, even if it is slow moving. And in the digital age, even when progress may seem to stall, new approaches can change things seemingly overnight. And even if the Word2Vec algorithm isn’t currently practical for Review, it may be useful during investigations or Early-Case-Assessment, where human reviewers simply need the extra insight AI can give them, allowing them to get to the facts of a case quicker, and then continue through the more standard processes required for defensibility.
Which is why it’s important that we continue to look outside the bubble of legaltech for possibilities. The answers are out there. We just need to make the connections.
Written by Jim Gill
Content Writer, Ipro
Stay up to date with the latest legaltech and eDiscovery trends in the Ipro Newsroom