Download PDF
In recent years, discovery in civil cases has been significantly impacted by advancements in technology, particularly with the rise of artificial intelligence and automation. As more firms adopt these technologies to improve the efficiency and quality of discovery processes, discovery is in for a radical transformation. This coming change seems poised to fundamentally transform the way we do discovery more so than even the advent of eDiscovery. In this blog post, we will explore the ways in which generative AI (“GAI”) will change civil discovery.
Increased quality of discovery output. While GAI has been known to hallucinate (make up something that does not exist), when used in a controlled and tested environment in GAI tools that are programmed for discovery processing, it is not difficult to imagine a discovery process in which GAI can produce higher quality output than that generated by human review. Consider a civil case where 50,000 documents must be reviewed to investigate and understand the exposure in a case and then release documents in production. Using eDiscovery platforms, a partner may divide the documents among a handful of associates to get the job done quickly. Those associates will flag problematic documents and report back on trends. This will take dozens of hours of work. A GAI tool can do this in a matter of minutes, and where as the associates may not have talked to each other at length and would have no way of knowing what trends might exist across their various sets of documents, the GAI tool can report back on trends seen in the entire set and make observations that human reviewers would have missed.
Increased quantity and quality of data in discovery. In many areas of law practice, collection and use of data is the issue of the future. In civil discovery, GAI tools are poised to introduce entirely novel approaches to data extraction, linking, summarization, and reasoning. This expansion goes beyond the current boundaries of predictive AI, marking a significant leap into new waters. The scope of GAI’s capabilities will not only encompass the analysis and identification of document characteristics during the discovery phase but will also extend to the generation of valuable insights and information about these documents. Metadata from documents, as gathered and stored by the original document created in the normal course of a job function, could also be analyzed. The layers of data and insights that can be drawn from them are difficult for the human mind to fathom, and nearly impossible for the limited capabilities of most humans to replicate with the consistency, speed, and accuracy that a computer can generate.
Increased complexity to document review. In some instances, the existence of such large quantities of data, especially data about the use of GAI itself, adds lawyers of complexity to the document review process. For example, consider a large set of e-mails, some of which were written with the assistance of a GAI tool like Co-Pilot, and some of which were written by a human alone, the old fashioned way. Assuming that there is data about the means used to write the e-mail or assuming that a GAI system would detect Co-Pilot assistance, when conducting document review it seems possible that it could be relevant to a case if a certain e-mail was written with the assistance of GAI or not. Whether that will be flagged, how it will be processed, and the means in which such GAI use will be analyzed are ongoing questions.
Limitations, too. While GAI has tremendous potential in the area of discovery, of course it has limitations. The technology is largely still experimental in some instances, with many questions remaining about how to use GAI in a manner consistent with our ethical duties as lawyers. And while GAI tools can catch things that humans miss, of course the inverse is also true. GAI doesn’t inherently understand emotions or cultural sensitivities. It might generate responses, or miss flagging content, that a human reviewer would recognize as insensitive, offensive, or inappropriate, especially in nuanced situations. Further, GAI tools tend to interpret language in a literal manner. They may struggle with understanding figurative language, sarcasm, or irony, leading to responses that may miss the intended humorous or nuanced meaning. While the technology improves and this risk will be mitigated more over time, it is not there yet, and it is unlikely that GAI can ever have the understanding that a human being brings.
Democratizing GAI: bridging the justice gap or widening the divide? As GAI becomes more available online for low or no cost, one of the biggest threats/opportunities in the law is whether this will level the playing field for pro se litigants and solo practitioners, or whether the implementation of sophisticated GAI systems by large firms and companies will conversely widen gaps in new ways. Or worse yet, create new gaps that previously did not exist. Already, large firms are developing GAI capabilities in-house by, for example, creating AI tools to help lawyers review and modify credit agreements and conduct due diligence. Still even more large firms have already announced the implementation of third-party GAI tools to perform substantive tasks such as legal research, document review, deposition preparation, and contract analysis. While it is too early to say how this question will ultimately be answered, it is quite possible that document heavy cases will be the proving ground from which the answer starts to emerge.
In conclusion, while the path ahead for GAI in the discovery domain is still unfolding, the transformative potential of this technology is undeniable. The coming years promise exciting developments as we unlock the complete capabilities of GAI, navigating through challenges and exploring new horizons in this dynamic field.
This content was generated in part by, and prepared for publication with the assistance of, ChatGPT.
