Scholarlink

<p>The complex and intricate features of legal language (referred to as <i>legalese</i>) have long been discussed in the literature, and scholars have advocated the use of plain language to enhance comprehension. The advent of ISO 24495-2 [<CitationRef CitationID="CR23">23</CitationRef>] on plain language and legal communication has brought about important changes in legal language drafting. According to the new standards, complex legal information must be conveyed clearly, effectively and accurately to become accessible to laypeople. In the current AI-driven mediascape, researchers have also experimented whether and how GenAI (more precisely, general chatbots) are capable of easifying legal convoluted language, albeit often reaching unconvincing results. This contribution is aimed at exploring whether current targeted GenAI tools (such as Notebook LM and ChatLaw) are able to disentangle the complexities of legal discourse, thus offering plain language alternatives. For this purpose, a host of contract clauses and statutory provisions are entered into the above-mentioned GenAI platforms, where the ISO 24495-2 is also uploaded for reference purposes. The chatbots are prompted to easify the texts inputted and re-write them in plain language by scanning and applying the principles expressed in the ISO standards. To allow for the evaluation of the AI-driven outputs, literature-driven easified versions of the legal extracts considered are accounted for. In this way, the scholarly recommended plain language versions of the texts are compared with the GenAI-powered outputs. The findings reveal very satisfactory results, with chatbots generating only a few minor inaccuracies. Namely, shortcomings mostly revolve around parenthetical information (which give rise to syntactical discontinuities), some non-simplified expressions (e.g., “unless the context suggests otherwise”), and the use of the passive form. In most of the cases, however, chatbot-based texts appear as more comprehensible and accessible than the literature-driven easified versions, thereby confirming the huge strides forward made by AI.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

How Well can GenAI Redraft Legalese into Plain Language? Insights into Legal Communication According to ISO 24495-2

  • 期刊论文

  • 发表日期：

    2026年7月17日

查看全文
  • Patrizia Giampieri

摘要

The complex and intricate features of legal language (referred to as legalese) have long been discussed in the literature, and scholars have advocated the use of plain language to enhance comprehension. The advent of ISO 24495-2 [23] on plain language and legal communication has brought about important changes in legal language drafting. According to the new standards, complex legal information must be conveyed clearly, effectively and accurately to become accessible to laypeople. In the current AI-driven mediascape, researchers have also experimented whether and how GenAI (more precisely, general chatbots) are capable of easifying legal convoluted language, albeit often reaching unconvincing results. This contribution is aimed at exploring whether current targeted GenAI tools (such as Notebook LM and ChatLaw) are able to disentangle the complexities of legal discourse, thus offering plain language alternatives. For this purpose, a host of contract clauses and statutory provisions are entered into the above-mentioned GenAI platforms, where the ISO 24495-2 is also uploaded for reference purposes. The chatbots are prompted to easify the texts inputted and re-write them in plain language by scanning and applying the principles expressed in the ISO standards. To allow for the evaluation of the AI-driven outputs, literature-driven easified versions of the legal extracts considered are accounted for. In this way, the scholarly recommended plain language versions of the texts are compared with the GenAI-powered outputs. The findings reveal very satisfactory results, with chatbots generating only a few minor inaccuracies. Namely, shortcomings mostly revolve around parenthetical information (which give rise to syntactical discontinuities), some non-simplified expressions (e.g., “unless the context suggests otherwise”), and the use of the passive form. In most of the cases, however, chatbot-based texts appear as more comprehensible and accessible than the literature-driven easified versions, thereby confirming the huge strides forward made by AI.