← 論文一覧に戻る

建築材料のエンボディドカーボン評価と脱炭素化のための説明可能・解釈可能な人工知能：PRISMAに基づく系統的・計量書誌学的レビュー

Explainable and interpretable artificial intelligence for embodied carbon assessment and decarbonization of building materials: A PRISMA-based systematic and bibliometric review (原題)

Funmilayo C. Jayeola, Masengo Ilunga

Asian Journal of Water, Environment and Pollution📚 査読済 / ジャーナル2026-10-16#AI×ESG経営インパクト: 調達リスク対象セクター: construction
DOI: 10.36922/ajwep026310212
原典: https://doi.org/10.36922/ajwep026310212
📄 PDF

🤖 gxceed AI 要約

日本語

本論文はPRISMA 2020に基づき2015〜2026年のScopus文献を系統的・計量書誌学的にレビューし、建築材料のエンボディドカーボン（EC）評価と脱炭素化におけるAI・説明可能AI（XAI）の活用動向を整理した。2023年以降研究が急増し、ニューラルネットやランダムフォレスト等がEC予測や材料最適化に用いられているが、XAI適用は限定的で予測精度に偏重している。地域データの空白やモデル移転性・実証検証の不足を指摘し、地域LCAデータ・AI予測・XAI解釈・最適化・意思決定支援を統合する概念フレームワークを提案する。

English

This PRISMA-based systematic and bibliometric review of Scopus literature (2015–2026) maps how AI and explainable AI (XAI) are used to assess embodied carbon (EC) and decarbonize building materials. Research surged after 2023, with ML methods (neural networks, random forests, SVM, gradient boosting, deep learning) applied to EC prediction and material optimization, yet XAI use remains limited and skewed toward accuracy over transferability. The authors flag regional data gaps and weak empirical validation, and propose a framework integrating localized LCA data, AI prediction, XAI interpretation, optimization, and decision support.

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

建設業はScope 3上流（資材）排出の比率が高く、SSBJ・有報でのScope 3開示やサプライヤーへのデータ要求が進む日本企業にとって、資材EC評価の透明性向上は実務課題。XAIによる説明可能性は第三者検証や監査対応の観点でも示唆に富む。

In the global GX context

Embodied carbon in materials is a major Scope 3 Category 1 driver under GHG Protocol and increasingly scrutinized in CSRD/ISSB building-sector disclosures. The review's call for explainable, validated AI models speaks directly to the transparency and assurance demands of global disclosure regimes.

👥 読者別の含意

🔬研究者:AI×LCA×XAIの研究ギャップ（地域データ・移転性・実証検証）と統合フレームワークを把握できる。

🏢実務担当者:資材EC評価にAI/XAIを導入する際の手法選択と透明性確保の指針が得られる。

🏛政策担当者:建設脱炭素政策においてAI活用の検証要件やデータ整備の必要性を示唆する。

📄 Abstract（原文）

Embodied carbon (EC) in building materials poses a significant challenge to decarbonizing the construction industry, necessitating data-driven methods for transparent and informed decisions. This study systematically reviews the use of artificial intelligence (AI) and explainable AI (XAI) in assessing EC and promoting low-carbon building materials. A PRISMA 2020–based review and bibliometric analysis of Scopus publications from 2015 to June 2026 explored trends, geographic distribution, research themes, AI/XAI methods, and their applications in EC assessment and material decarbonization. Research has notably increased in 2023, highlighting AI’s growing role in life cycle assessments and material decarbonization. Machine learning methods such as neural networks, random forests, support vector machines, gradient boosting, and deep learning are increasingly used for EC prediction, material performance evaluation, and optimization. However, XAI applications are limited, with more focus on prediction accuracy than model transferability. Key XAI techniques include Shapley Additive exPlanations, partial dependence plots, sensitivity analysis, and feature importance analysis. The review identifies limitations such as regional data gaps, limited model explainability, limited database integration, and limited empirical validation. A conceptual framework is proposed combining localized life cycle data, AI prediction, XAI interpretation, optimization, and decision support. The review concludes that XAI can enhance transparency and aid low-carbon material decisions, though further validation across laboratories, manufacturing, and regions is needed for practical implementation.

🔗 Provenance — このレコードを発見したソース

🔔 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。