Dataset for: Artificial Intelligence and ESG Disclosure: Evaluating Large Language Models for Identifying Symbolic and Substantive Sustainability Communication
人工知能とESG開示：シンボリックおよび実質的なサステナビリティコミュニケーションの識別における大規模言語モデルの評価 (AI 翻訳)
Anuj Pal
🤖 gxceed AI 要約
日本語
本研究は、大規模言語モデル（LLM）がESG開示におけるシンボリック（表面的）なコミュニケーションと実質的なコミュニケーションを識別できるかを評価する。公開企業の年次・サステナビリティ報告書から抽出したESG開示データセットを構築し、LLMの性能を検証した。結果は、LLMが実質的な開示を高い精度で識別できることを示す一方、シンボリックな開示の識別には課題が残る。
English
This study evaluates the ability of large language models (LLMs) to distinguish between symbolic and substantive sustainability communication in ESG disclosures. It constructs a dataset from annual and sustainability reports of publicly listed companies and tests LLM performance. Results show LLMs can identify substantive disclosures with high accuracy but struggle with symbolic ones, highlighting both promise and limitations in automated ESG analysis.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
日本ではSSBJ基準の導入が進み、ESG開示の質的評価が重要視されている。本論文はLLMを用いて開示の実質性を評価する手法を提供し、日本の投資家や企業がシンボリックな開示（いわゆる「グリーンウォッシュ」）を見極めるためのツール開発に示唆を与える。また、有報や統合報告書の分析自動化にも応用可能である。
In the global GX context
Globally, as ISSB and CSRD push for more rigorous sustainability disclosures, the ability to distinguish substantive from symbolic communication becomes critical. This paper advances the use of AI for automated ESG analysis, offering a framework that can enhance greenwashing detection and improve disclosure quality assessment across markets. It provides a reproducible dataset that can serve as a benchmark for future AI-based sustainability text analysis.
👥 読者別の含意
🔬研究者:Provides a benchmark dataset and evaluation framework for using LLMs in ESG disclosure analysis, advancing AI×ESG research.
🏢実務担当者:Offers a tool to assess the authenticity of sustainability communications, aiding in greenwashing risk management and disclosure improvement.
🏛政策担当者:Highlights the potential of AI to monitor and enforce substantive disclosure, informing regulatory oversight and standard-setting.
📄 Abstract（原文）
Artificial Intelligence and ESG Disclosure:Evaluating Large Language Models for Identifying Symbolic and Substantive Sustainability Communication Contents: - esg_coded_dataset.xlsx - Annual and sustainability reports used for ESG disclosure coding. The reports were obtained from the official investor relations or sustainability webpages of the respective companies and were used as source material for the research. This dataset accompanies the manuscript Artificial Intelligence and ESG Disclosure: Evaluating Large Language Models for Identifying Symbolic and Substantive Sustainability Communication. It contains the ESG-coded dataset used in the study and the publicly available corporate reports that served as the source documents. ESG disclosure excerpts were extracted from these reports, systematically coded using the Operational ESG Disclosure Coding Framework (OEDCF), and analyzed to evaluate the performance of large language models in distinguishing symbolic and substantive sustainability communication. This repository is provided to promote transparency and facilitate the reproducibility of the research.
🔗 Provenance — このレコードを発見したソース
- openalex https://doi.org/10.5281/zenodo.21509495first seen 2026-07-26 05:35:12 · last seen 2026-07-26 05:35:13
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。