跳到主要內容

簡易檢索 / 詳目顯示

研究生: 劉映廷
Liu, Ying-Ting
論文名稱: 生成式人工智慧於司法文書撰擬之應用界限
The Limits of Generative AI in the Drafting of Judicial Documents
指導教授: 莊弘鈺
口試委員: 鄭婷嫻
鄭菀瓊
學位類別: 碩士
Master
系所名稱: 商學院 - 科技管理與智慧財產研究所
Graduate Institute of Technology, Innovation and Intellectual Property Management
論文出版年: 2026
畢業學年度: 114
語文別: 中文
論文頁數: 114
中文關鍵詞: 生成式人工智慧司法文書判決書強制執行文書正當法律程序司法獨立紅綠燈模式
外文關鍵詞: Generative Artificial Intelligence, Judicial Documents, Judgments, Compulsory Enforcement Documents, Due Process, Judicial Independence, Traffic Light Model
相關次數: 點閱:10下載:0
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報
  • 隨著生成式人工智慧技術之興起,司法從業人員得否借助此一技術減輕文書負擔,已成為司法制度設計之重要課題。本文以我國法院案件量持續攀升、文書工作高度密集為研究背景,聚焦於生成式AI得否介入司法文書撰擬,以及不同類型司法文書之應用界限應如何劃定之探討。
    就研究方法而言,本文採取比較法分析途徑,以美國作為主要觀察對象,並將研究範圍限縮於「判決書」與「強制執行文書」,蓋此二者分居司法文書光譜之兩端,前者涉及裁判權核心,後者則呈現高度程序性與格式化特徵。
    本文首先析論生成式人工智慧之技術特性,指出大型語言模型之機率生成本質與法律確定性要求之間存在根本性扞格,幻覺現象之系統性風險以及黑盒不可解釋性,均對正當法律程序與裁判理由說明義務構成制度挑戰。其次,本文就美國聯邦司法行政體系之治理架構、聯邦證據法規範,以及加州、紐約州、賓夕法尼亞州等州之具體規範加以比較分析。再者,本文就台灣現行法制進行審查,檢視人工智慧基本法、司法院使用指引及司法院裁判書草稿自動生成系統之暫緩等發展,並以司法官訓練人員及法官助理之訪談作為實務佐證。
    在比較分析基礎上,本文提出以文書對審判核心之涉入程度作為分級依據之「紅綠燈模式」:紅燈區絕對禁止生成式人工智慧介入,涵蓋判決書之心證形成、事實認定及法律適用等核心環節;黃燈區有限度容許,須具備強制人工覆核、輸出可追溯性及禁止輸入非公開資料三項控管機制;綠燈區原則允許,適用於高度格式化之例行性強制執行文書,惟須保留人類最終確認程序。


    With the rise of generative artificial intelligence, the question of whether judicial practitioners may leverage this technology to alleviate document-drafting burdens has become a significant issue in judicial system design. Against the backdrop of steadily increasing caseloads and highly document-intensive workflows in Taiwan's courts, this article focuses on two central questions: whether GenAI may intervene in the drafting of judicial documents, and how the limits of its application should be defined across different document types.
    In terms of methodology, this article adopts a comparative law approach, with the United States as the primary reference, and confines its scope to "judgments" and "compulsory enforcement documents." These two types occupy opposite ends of the judicial document spectrum, as the former concerns the core of adjudicative power while the latter is characterized by highly procedural and standardized content.
    The article begins by analyzing the technical characteristics of generative AI, arguing that the probabilistic generative nature of large language models is fundamentally at odds with the demands of legal certainty, and that the systemic risks of hallucination and the opacity of black-box models pose institutional challenges to due process and the judicial duty to give reasons. It then undertakes a comparative analysis of the governance framework of the U.S. federal judicial administration, federal evidence law, and the specific regulations of states including California, New York, and Pennsylvania. The article further examines Taiwan's current legal landscape, reviewing developments such as the Artificial Intelligence Basic Act, the Judicial Yuan's usage guidelines, and the suspension of the Judicial Yuan's automated judgment-drafting system, supplemented by interviews with judicial trainees and law clerks as practical evidence.
    Drawing on this comparative analysis, the article proposes a "Traffic Light Model" that classifies permissibility based on the degree to which a document implicates the core of adjudication: the red zone absolutely prohibits GenAI intervention, covering core aspects of judgments such as the formation of judicial conviction, fact-finding, and legal application; the yellow zone permits limited use, subject to three control mechanisms of mandatory human review, output traceability, and a prohibition on inputting non-public case materials; and the green zone generally permits use for highly standardized, routine compulsory enforcement documents, provided that a human final-confirmation procedure is retained.

    第一章、緒論 1
    第一節、研究動機與目的 1
    第二節、研究問題 4
    第三節、研究方法與範圍 5
    第二章、生成式人工智慧介入司法文書之理論基礎與爭議 7
    第一節、生成式人工智慧之技術特性與挑戰 7
    第二節、既有文獻回顧 13
    第三節、司法文書之功能與類型 21
    第三章、美國法制與司法實務 28
    第一節、司法體系中人工智慧之規範架構與基本原則 28
    第二節、人工智慧介入司法裁判之憲法界限 34
    第三節、聯邦法院層級對人工智慧使用之回應 40
    第四節、州法院層級對人工智慧使用之規範 63
    第五節、綜整分析 76
    第四章、我國法制現況與比較法觀察 79
    第一節、台灣法制與實務現況 79
    第二節、台、美法制比較 89
    第三節、對我國司法實務之建議 95
    第五章、結論 105
    參考文獻 107

    一、 中文文獻
    書籍
    沈建興,強制執行法逐條釋義(上)初版第一刷,元照出版有限公司,2014年。
    沈建興,強制執行法逐條釋義(下),二版,元照出版有限公司,2020年。
    張登科,強制執行法修訂版,臺北市:三民書局總經銷,2025年。
    許士宦,口述講義民事訴訟法(下),增訂一版,臺北市:新學林,2019年。
    行政資料與政府報告
    司法院,司法統計年報,2024年。
    司法院院台廳司二字第1150500116號函
    司法院院台廳司二字第1150500529號函。
    立法院公報96卷48期(二冊),2007年。
    期刊論文
    王道維、邱筱涵,當ChatGPT來敲法官的門—淺談AI應用於司法審判的原則與方式,當代法律,第18期,頁6-27,2023年6月。
    林勤富,智慧法院之發展與界限(上)──演算法、科技治理與司法韌性,月旦法學雜誌,第323期,頁72-98,2022年4月。
    林勤富,智慧法院之發展與界限(下)──演算法、科技治理與司法韌性,月旦法學雜誌,第324期,頁109-131,2022年5月。
    邱文聰,智慧政府資料驅動決策的法治挑戰以「模仿型」及「相關性發掘型」人工智慧系統為例,中研院法學期刊,第37期,頁27-97,2025年9月。
    邱惠美,我國有關司法事務官處理程序之探討-兼論我國消費者保護官制度,消費者保護研究,第12輯,頁153-209,2006年12月。
    姜世明,概論民事訴訟中自由心證主義之內涵與界限-研討會引言,台灣法學雜誌,第196期,頁36-51,2012年。
    張巧函,生成式AI作為司法人員決策輔助工具的可能性初探:以ChatGPT為例,文官制度,第16卷第2期,頁97-137,2024年。
    陳淑芳,人工智慧在公部門之應用與國家損害賠償責任,世新法學,第十七卷第一號,2023年。
    陳豐年,淺論美國司法領域的人工智慧應用對臺灣的啟示,月旦律評,第41期,頁85-99,2025年8月。
    曾小榕、金孟華,AI在刑事證據法之研究:以美國聯邦證據規則為核心,歐美研究,第55卷第1期,頁1-60,2025年。
    黃詩淳,AI可解釋性的法學意義及其實踐,國立臺灣大學法學論叢,第52卷特刊,頁931-972,2023年。
    楊仁壽,法學方法論,日新,第7期,頁19-28,2006年11月。
    廖建瑜,AI在法院運用與界限,司法周刊,第2310期,頁19-28,2026年5月。
    蔡立亭,第三審法官裁判之詮釋與規範-以民事訴訟法第469條第6款為核心,華岡法粹,第63期,頁131-154,2017年。
    龍建宇、莊弘鈺,人工智慧於司法實務之可能運用與挑戰,中正大學法學集刊,62期,頁43-108,2019年。
    蘇凱平,以人工智慧輔佐法院之刑罰裁量:破譯「黑盒子」的可能與顧慮,臺大法學論叢,第54卷第3期,頁875-955,2025年。
    博碩士學位論文
    許淑媛,AI在法律執行之應用與隱私權保護之研究,國立中正大學法律學系博士論文,頁1-185,2024年。
    網際網路
    Chris,公務員想用ChatGPT請注意了!國家級「AI準則」有什麼地雷要避,INSIDE硬塞的網路趨勢觀察,2023年8月31日,https://www.inside.com.tw/article/32679-taiwan-government-using-chatgpt-guide(最後瀏覽日期:2026年6月4日)。
    司法院,因應國民法官新制,司法院啟用AI量刑資訊系統,2023年2月6日,https://www.judicial.gov.tw/tw/cp-1887-806741-d6471-1.html(最後瀏覽日期:2026年6月5日)。
    司法院,司法院審慎發展生成式AI應用,以撰寫刑事裁判草稿初試啼聲;期望減輕法官工作負荷,審判核心仍由法官自行決定,2023年8月27日,https://www.judicial.gov.tw/tw/cp-1887-929494-8a9fb-1.html(最後瀏覽日期:2026年6月5日)。
    民間司法改革基金會,AI草擬判決的三大疑問與三大風險 要減輕負擔,也要控制AI風險,2023年9月26日,https://www.jrf.org.tw/articles/2550(最後瀏覽日期:2026年6月5日)。
    司法院,司法院就民間團體2023年9月26日召開「AI草擬判決的三大疑問與三大風險」記者會所提建議之回應新聞稿,2023年9月27日,https://www.judicial.gov.tw/tw/cp-1887-951341-9add3-1.html(最後瀏覽日期:2026年6月5日)。
    司法院,回應各界對於「智慧化裁判草稿系統」之疑慮—司法院:刻正研擬參考指引、審慎評估上線期程,2023年10月6日,https://www.judicial.gov.tw/tw/cp-1887-956786-3f78a-1.htm(最後瀏覽日期:2026年6月5日)
    維基百科,〈檢索增強生成〉,維基百科,2026年4月18日,https://zh.wikipedia.org/zh-tw/%E6%AA%A2%E7%B4%A2%E5%A2%9E%E5%BC%B7%E7%94%9F%E6%88%90(最後瀏覽日期:2026/06/03)。
    數位發展部,AI風險分類框架,2026年5月28日更新,https://moda.gov.tw/major-policies/ai/governance/19244(最後瀏覽日期:2026年6月4日)。
    二、 英文文獻
    判決
    Carpenter v. United States, 585 U.S. 296 (2018).
    Chambers v. NASCO, Inc., 501 U.S. 32 (1991).
    Commodity Futures Trading Comm’n v. Schor, 478 U.S. 833 (1986).
    County of Sacramento v. Lewis, 523 U.S. 833 (1998).
    Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 (1993).
    Goldberg v. Kelly, 397 U.S. 254 (1970).
    Link v. Wabash R. Co., 370 U.S. 626 (1962).
    Mathews v. Eldridge, 424 U.S. 319 (1976).
    Northern Pipeline Constr. Co. v. Marathon Pipe Line Co., 458 U.S. 50 (1982).
    Ross v. United States, 331 A.3d 220 (D.C. 2025).
    Snell v. United Specialty Insurance Co., 102 F.4th 1208 (11th Cir. 2024).
    State v. Loomis, 881 N.W.2d 749 (Wis. 2016).
    Stern v. Marshall, 564 U.S. 462 (2011).
    United States v. Jones, 565 U.S. 400 (2012).
    United States v. Raddatz, 447 U.S. 667 (1980).
    Whalen v. Roe, 429 U.S. 589 (1977).
    Withrow v. Larkin, 421 U.S. 35 (1975).
    行政資料與政府報告
    ADV. COMM. ON AI & THE CTS., Annual Report To The Chief Judge And Chief Administrative Judge Of The State Of New York (Dec. 2025), https://www.nycourts.gov/LegacyPDFS/publications/pdfs/AI%20Annual%20Report%202025.pdf (last visited June 12, 2026).
    Allison H. Goddard, Chambers Guidelines for Use of Generative AI (Aug. 12, 2025), https://www.ali.org/special-events/genai-chambers-webinar (last visited June 12, 2026).
    Andrea L. Miller et al., AI Readiness for the State Courts (NAT'L CTR. FOR STATE COURTS, Sept. 2025), https://www.ncsc.org/sites/default/files/media/document/AI_Readiness_for_the_State_Courts_2025.pdf (last visited June 12, 2026).
    BOLCH JUD. INST. & AM. L. INST., Using Generative AI in Chambers (Nov. 2025), https://www.ali.org/special-events/genai-chambers-webinar (last visited June 12, 2026).
    COMM. ON RULES OF PRAC. & PROC. OF THE JUD. CONF. OF THE U.S., Agenda Book (June 10, 2025), https://www.uscourts.gov/forms-rules/records-rules-committees/agenda-books/committee-rules-practice-and-procedure-june-2025 (last visited June 12, 2026).
    Grassley, Charles E., Chairman, S. COMM. ON THE JUDICIARY, Letter to Henry T. Wingate, U.S. Dist. Judge for the S. Dist. of Miss. (Oct. 6, 2025), https://www.grassley.senate.gov/imo/media/doc/grassley_to_the_honorable_henry_twingateaioversight.pdf (last visited June 12, 2026).
    James E. Baker, Laurie N. Hobart & Matthew Mittelsteadt, An Introduction to Artificial Intelligence for Federal Judges, FED. JUD. CTR. (2023), https://www.fjc.gov/sites/default/files/materials/47/An_Introduction_to_Artificial_Intelligence_for_Federal_Judges.pdf (last visited June 12, 2026).
    JUDICIAL COUNCIL OF CAL., Report to the Judicial Council, Item No. 25-109: Judicial Branch Administration: Rule and Standard for Use of Generative Artificial Intelligence in Court-Related Work 1 (June 16, 2025).
    Letter from Charles E. Grassley, Chairman, S. COMM. ON THE JUDICIARY, to Julien Xavier Neals, U.S. Dist. Judge for the Dist. of N.J. (Oct. 6, 2025), https://www.grassley.senate.gov/imo/media/doc/grassley_to_the_honorable_julien_xavier_neals_-_ai_oversight.pdf (last visited June 12, 2026).
    Letter from Julien X. Neals, U.S. Dist. Judge for the Dist. of N.J., to Robert J. Conrad, Jr., Dir., ADMIN. OFFICE OF THE U.S. COURTS 1 (Oct. 20, 2025), https://www.grassley.senate.gov/imo/media/doc/neals_to_grassley_re_judiciary_use_of_ai.pdf (last visited June 12, 2026).
    Letter from Robert J. Conrad, Jr., DIR., ADMIN. OFF. U.S. CTS., to Charles W. Grassley, CHAIRMAN, S. COMM. ON THE JUDICIARY (Oct. 21, 2025), https://www.grassley.senate.gov/imo/media/doc/ao_to_grassley_re_judiciary_use_of_ai.pdf (last visited June 12, 2026).
    Letter from Henry T. Wingate, U.S. Dist. Judge for the S. Dist. of Miss., to Robert J. Conrad, Jr., Dir., ADMIN. OFFICE OF THE U.S. COURTS 1 (Oct. 21, 2025), https://www.grassley.senate.gov/imo/media/doc/wingate_to_grassley_re_judiciary_use_of_ai.pdf (last visited June 12, 2026).
    N.Y. STATE UNIFIED COURT SYS., Interim Policy on the Use of Artificial Intelligence (Oct. 2025), https://www.nycourts.gov/LegacyPDFS/a.i.-policy.pdf (last visited June 12, 2026).
    NAT’L CTR. FOR STATE CTS., Principles and Practices for Using AI Responsibly and Effectively in State Courts 2-3 (2025), https://www.ncsc.org/resources-courts/principles-practices-ai-use-courts (last visited June 12, 2026).
    PA. JOINT JUDICIAL ETHICS ADVISORY BD., General Ethics Guidance No. 2-2025: Use of Generative Artificial Intelligence (Dec. 10, 2025), https://www.pacourts.us/Storage/media/pdfs/20251215/201356-generalguidanceno.2-2025.pdf (last visited June 12, 2026).
    PA. UNIFIED JUDICIAL SYS., Interim Policy on the Use of Generative Artificial Intelligence by Judicial Officers and Court Personnel (2025), https://share.google/r0W6bq8dKioNk9eNx(last visited June 12, 2026).
    ADMIN. OFF. U.S. CTS., Profile of the Administrative Office of the U.S. Courts: Annual Report 2013 (2013), https://www.uscourts.gov/data-news/reports/annual-reports/directors-annual-report/annual-report-2013/profile-administrative-office-us-courts-annual-report-2013.
    Scott Schlegel, AI in Chambers: A Framework for Judicial AI Use (Sept. 2025), https://static1.squarespace.com/static/63c35a0b6817d035ea368c19/t/68c87476f545721848a413d1/1757967478473/AI_in_Chambers_v1_1_Judge+Schlegel.pdf (last visited June 12, 2026).
    ADVISORY COMM. ON EVIDENCE RULES, Agenda Book (Oct. 27, 2023), https://www.uscourts.gov/forms-rules/records-rules-committees/agenda-books/advisory-committee-evidence-rules-october-2023 (last visited June 12, 2026).
    ADVISORY COMM. ON EVIDENCE RULES, Agenda Book (May 7, 2026), https://www.uscourts.gov/forms-rules/records-rules-committees/agenda-books/advisory-committee-evidence-rules-may-2026 (last visited June 12, 2026).
    期刊
    A.D. (Dory) Reiling, Courts and Artificial Intelligence, 11 INT'L J. COURT ADMIN. 1 (2020).
    Anika Jaitley, Daniel W. Linna Jr., Hon. Xavier Rodriguez, V.S. Subrahmanian & Siyu Tao, Artificial Intelligence in Federal Courts: A Random-Sample Survey of Judges, 27 SEDONA CONF. J. (forthcoming 2026).
    Brandon Waldon et al., Large Language Models for Legal Interpretation? Don't Take Their Word for It, 114 GEO. L.J. 115 (2025).
    Daniel J. Capra, Deepfakes Reach the Advisory Committee on Evidence Rules, 92 FORDHAM L. REV. 2491 (2024).
    Daniel J. Solove, Access and Aggregation: Privacy, Public Records, and the Constitution, 86 MINN. L. REV. 1137 (2002).
    Daniel Martin Katz, Michael James Bommarito, Shang Gao & Pablo Arredondo, GPT-4 Passes the Bar Exam, 382 PHIL. TRANS. R. SOC. A 20230254 (2024).
    Danielle Keats Citron, Technological Due Process, 85 WASH. U. L. REV. 1249 (2008).
    David Uriel Socol de la Osa & Nydia Remolina, Artificial Intelligence at the Bench: Legal and Ethical Challenges of Informing—or Misinforming—Judicial Decision-Making Through Generative AI, 6 DATA & POL'Y, e59 (2024).
    Dovilė Barysė & Roee Sarel, Algorithms in the Court: Does it Matter Which Part of the Judicial Decision-Making is Automated?, 32 ARTIFICIAL INTELLIGENCE AND LAW 117 (2024).
    Harry Surden, Artificial Intelligence and Law: An Overview, 35 GA. ST. U. L. REV. 1305 (2019).
    Matthew Dahl, Varun Magesh, Mirac Suzgun & Daniel E. Ho, Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models, 16 J. LEGAL ANALYSIS 64 (2024).
    Stephanie A. J. Dias & Renato M. Sátiro, Artificial Intelligence in the Judiciary: A Critical View, 164 FUTURES art. 103493 (2024).
    Victoria Hendrickx, Rethinking the Judicial Duty to State Reasons in the Age of Automation?, 1 CAMBRIDGE F. ON AI: L. & GOVERNANCE e26 (2025).
    W. Bradley Wendel, The Promise and Limitations of Artificial Intelligence in the Practice of Law, 72 OKLA. L. REV. 21 (2019).
    Xavier Rodriguez, Judging AI: How U.S. Judges Can Harness Generative AI Without Compromising Justice, 109 JUDICATURE 10 (2025).
    網際網路
    Antonio Virgili, The Role of Artificial Intelligence in the Judiciary System: An International Perspective, CENTRO STUDI INT'L 3 (2025), https://www.academia.edu/127934396/The_role_of_Artificial_Intelligence_in_the_judiciary_system_an_international_perspective (last visited Apr. 15, 2026).
    California Courts Adopt Generative AI Policies, L.A. MAG. (Aug. 21, 2025), https://lamag.com/news/landmark-ruling-california-courts-adopt-generative-ai-policies/ (last visited June 12, 2026).
    Judicial Conference of the U.S., Governance of the Judiciary, U.S. Courts, https://www.uscourts.gov/administration-policies/governance-judicial-conference (last visited Apr. 15, 2026).
    NAT'L CTR. FOR STATE CTS., About Us, NCSC, https://www.ncsc.org/about-us (last visited May 27, 2026).
    NAT'L CTR. FOR STATE CTS., Judicial Use of Generative AI: Lessons Learned, NCSC, https://www.ncsc.org/resources-courts/judicial-use-generative-ai-lessons-learned (last visited June 12, 2026).
    Oliver Roberts, Legal Regards with Oliver Roberts, Judge Allison Goddard on AI Use by the Judiciary, Judicial Discretion and Decision-Making, Legal Regards With Oliver Roberts, YouTube (Mar. 7, 2026), https://www.youtube.com/watch?v=KUptLpGRybs (last visited June 12, 2026).
    Walsworth LLP, New 2026 Court Rules Revamp California Civil Litigation, Walsworth Law (Dec. 16, 2025), https://walsworthlaw.com/news-and-insights/new-2026-court-rules-revamp-california-civil-litigation/ (last visited June 12, 2026).

    QR CODE
    :::