Since its inception, TAR has been met with considerable skepticism, and this skepticism was not unfounded. In early stages, TAR proved to be a laborious process, demanding users to invest more time in training and quality control than was expected or available. Moreover, there was no guarantee of the overall performance even after all the time invested. Consequently, it faced widespread rejection. However, as we fast forward to the present, we have witnessed remarkable advancements in AI technology and workflows that achieve successful and repeatable outcomes. Therefore, now is the opportune time to reassess CAL and capitalize from the improvements.
Let us deconstruct the current CAL workflow and recognize why embracing and adopting it should be an unquestionable choice.
Essentially, CAL employs supervised learning to grasp how reviewers classify documents as "good" or "bad". For example, it focuses on identifying documents with specific themes based on what is tagged as either "responsive" or "non-responsive". By leveraging the patterns observed in reviewers' tagging patterns, CAL arranges the documents in order of their responsiveness, thus expediting the review process.
In recent years, most content and social media platforms have embraced a comparable concept through their "explore pages" or “like” inputs. Applications such as Instagram, YouTube, and Pinterest learn about users' preferences and behaviors, presenting the most relevant and captivating posts without requiring explicit search terms. As you continue scrolling, less relevant posts gradually appear until they dominate your feed. CAL adopts a similar approach, prioritizing responsive and non-responsive documents based on the reviewer's behavior, ensuring that the most pertinent information takes precedence.
A common misconception suggests that TAR technology is prone to misplacing documents or identifying documents for production without human supervision, resulting in apprehension. However, this notion is incorrect. CAL actually organizes documents in a manner that presents irrelevant documents last, but it is ultimately up to the case team to determine their fate. Thus, the quality of CAL review is a combination of the technology and the expertise and precision of the reviewer in tagging documents.
During the Meet and Confer process, both parties establish agreed-upon criteria for when to cease document review based on the technology's accuracy. This means that any decision to stop reviewing documents is mutually predetermined. The case team always maintains the final authority and control.
While it's possible that a few responsive documents may be overlooked in a dataset of 500,000, the question arises whether it is worth the time and resources to locate those documents. The likelihood of those documents being the crucial evidence is minimal. Hence, the final decision centers around striking a balance between meeting RFP requirements while avoiding unnecessary financial or time constraints for documents that are unlikely to contain relevant information.
Considering the significant time and cost savings achieved using CAL, it is surprising that some practitioners still reject its implementation. Our Fusion product ensures that the most experienced reviewers utilize the tool to its fullest potential, resulting in time, cost, and stress reduction for you.