P2P

22 P E E R T O P E E R : I L T A ' S Q U A R T E R L Y M A G A Z I N E | S U M M E R 2 0 2 2 have an influence within the next one to five years. Typically there will be particular applications with material impact, and those should be leveraged. There will then be other areas where it makes sense to tread more carefully and apply the technology only to the extent that it can help. In the field of law, the areas of material and possibly transformative impact for AI in the near term are e-discovery for litigators and, more generally, content analysis and content management. For example, AI can generate routine legal documents such as motions and pleadings – compiling answers and responses in a matter of minutes and freeing up a lawyer's precious time to do more critical litigation tasks. Regarding the application of AI to speech recognition, it will certainly help the current transcription process, but it won't change everything. AI-powered speech recognition has particular usefulness in controlled single-participant settings, such as legal dictation and note management, but has limited usefulness for things like the capture and transcription of records of legal proceedings such as depositions and hearings. Let's go into that in a little more detail. AI in Legal Proceedings Legal proceedings such as depositions, arbitration hearings and examinations under oath represent particular challenges with regard to effectively and accurately capturing the record. The official capturing the record has two main responsibilities. First, they must ensure accurate and unbiased procedure such that no party to the case can unduly influence the records of testimony or of any other discourse. They have specific facilitation responsibilities in that regard, including such duties as identifying and taking affirmations of witnesses, marking and distributing associated evidentiary artifacts such as document exhibits and ensuring that off-record discussion is not captured and transcribed for distribution. Second, they must ensure the accurate capture and ultimate transcription of the record. While AI-powered speech recognition can be useful in the rough transcription of audio and other forms of capture, it cannot approach the accuracy of a trained reporter who annotates and clarifies the captured record such that a highly accurate transcript can be produced. Whereas machine-learning algorithms have capabilities human beings don't – for example, going through decades of data for useful insights – such capabilities are not applicable in the context of human interactions in legal proceedings because a) the context of proceedings can be random and quite chaotic, so there is no opportunity for the machine to learn the particular jargon, taxonomy or subject matter that may be discussed in any particular setting; b) the machine cannot be effectively trained on the particular speech idiosyncrasies of the participants such as accent, style and mannerism; and c) the audio quality varies dramatically from setting to setting and even from participant to participant within a particular setting. Considering all of those challenges together, it's not surprising that AI-powered automatic speech recognition in the legal proceeding setting ranges wildly in accuracy and almost never meets the requirements of a traditional "rough" transcript of a proceeding without the applied effort of a transcriptionist, proofer or scopist reviewing and correcting the draft automatic transcription. Aside from the technical limitations of AI-powered speech recognition, there is still the matter of the reporter's officiating and facilitation responsibilities. In most legal proceedings the duties assigned to the reporter are critical to ensuring the validity and credibility of the record. Technology cannot and will not replace the assurance that an unbiased party has assured the accurate capture and transcription of the proceeding, and it cannot replace the affirmation and document management and distribution duties. F E A T U R E S

• Cover