1. NCCU Academic Hub
  2. Publications
  3. College of Social Science
  4. Theses
  5. Item

Publications-Theses

Article View/Open

Publication Export

Google ScholarTM

NCCU Library

Citation Infomation

Related Publications in TAIR

題名 人資與善的距離？勞工績效改善計畫之研究 - 以閉源型生成式AI驗證我國法院判決
How Far is HR from "Goodness"? A Study of Performance Improvement Plans for Incompetent Employees Using Closed - Source Generative AI to Analyze Taiwanese Court Judgments
作者 陳詩吟
Chen, Shih-Yin
貢獻者 劉梅君
Liu, Mei-Chun
陳詩吟
Chen, Shih-Yin
關鍵詞 績效改善計畫
PIP
S.M.A.R.T.
漸進式懲戒
閉源型生成式AI
performance improvement plan
PIP
S.M.A.R.T.
progressive discipline
closed-source generative AI
日期 2025
上傳時間 4-Aug-2025 15:42:17 (UTC+8)
摘要 　　本文題目「人資與善的距離」，以人資依據解僱最後手段性原則的觀點，思考企業面對不能勝任工作的勞工時，究竟採取多少善意提供勞工改善機會？由於，我國面臨中高齡化的勞動力市場，企業存在勞工高離職率，本研究思考如何留任在職勞工，茲從法理解釋、法院判決與學者論述，觀察前人研究無法改變勞工被資遣，進而關注「績效改善計畫」（簡稱：PIP）輔導不能勝任工作之勞工。企業重視法官的PIP見解，另發現神經網路能將文字轉成數字進行演算，冀望人資不倉促資遣勞工，運用人工智慧（簡稱：AI）分類判決並制定PIP。 　　本研究從收集西文期刊彙整PIP制定過程，另從司法院裁判系統檢索判決全文，關鍵字為PIP、績效改善、績效改進、績效輔導等任何一項，裁判期間設定截至2025年03月31日共498篇PIP判決。其次，採取ChatGPT-4o、Gemini 2.5 Pro、Claude 3.7與Grok 3等四家閉源型生成式AI分類PIP判決，執行描述性統計、分類準確度與卡方獨立性檢定，嘗試找出準確度相對較佳的AI工具；再者，歸納法官肯認企業對不能勝任勞工採取PIP措施且判決資方勝訴的要件。 　　研究發現，西文期刊與我國法院判決在PIP皆運用S.M.A.R.T管理不能勝任勞工，歸納S具體的指「佐證客觀疏失與評估主觀疏失」、M可測量的指「具體指標與量化或順位」、A可達成的指「比較勞工及同事」、R相關的指「達到僱用之客觀合理經濟目的」，T時效的指「改善、追蹤與通知，普遍輔導三個月」；此外，人資在PIP實施漸進式懲戒應落實企業內部辦法，訪談過程採取尊重勞工措施，總之，人資妥善制定PIP能夠降低資方訴訟敗訴，建議人資在PIP訪談前充分瞭解勞工個案且訪談時察言觀色。其次，在AI運用檢索增強生成（簡稱：RAG）有助於分類法院判決，Claude 3.7分類判決準確度為92%與83%，高於ChatGPT-4o準確度為87%與61%，因此，閉源型生成式AI達到快速分類PIP判決，建議人資學習AI提問技巧，遇到勞資糾紛時有能力使用AI整理法官見解。
　　This study, from a Human Resources (HR) perspective, examines the extent of goodwill and opportunities for improvement that companies should offer to employees deemed incompetent, especially when considering the ultima ratio doctrine of dismissal. Given Taiwan’s aging workforce and high employee turnover rates, this study focuses on strategies for employee retention. While previous legal interpretations, court judgments, and academic literature haven't fully succeeded in preventing terminations, this research highlights the critical role of Performance Improvement Plan(PIP) to guide employees who are incompetent. Recognizing the growing importance of judicial interpretations of PIP for businesses, and the potential of neural networks to process legal texts, the present study advocates for HR professionals to move beyond hasty terminations by leveraging Artificial Intelligence (AI) to classify court judgments and formulate more effective PIP. 　　This study first reviewed the process of designing PIP based on western academic literature. Subsequently, 498 court judgments concerning PIP were gathered from Judicial Yuan’s Judgment System up to March 31, 2025, using keywords such as “PIP”, “performance improvement”, or “performance coaching”. Furthermore, four closed-source generative AI models (i.e., ChatGPT-4o, Gemini 2.5 Pro, Claude 3.7, and Grok 3) were employed to classify these PIP judgments. Descriptive statistics, classification accuracy analysis, and chi-square tests of independence were then performed to identify the AI tool with relatively superior classification accuracy for legal texts. Finally, the judicial arguments that recognize the adoption of PIP by employers as a significant factor for incompetent were also summarized. 　　The findings indicate that both Western academic literature and Taiwanese court judgments apply the S.M.A.R.T. principle when managing incompetent employees through PIP. S (Specific) involves “substantiating objective misconduct and evaluating subjective deficiencies.” M (Measurable) entails “setting concrete indicators and quantifying performance, or ranking.” A (Attainable) means “comparing the employee’s performance with that of colleagues.” R (Relevant) represents “achieving the objective and reasonable economic purpose of employment.” Lastly, T (Timely) denotes “improvement, monitoring, and notification, typically involving a three-month coaching period.” Beyond the S.M.A.R.T. framework, HR professionals should implement PIP in alignment with the principle of progressive discipline and ensure the enforcement of internal company policies, while adopting a respectful approach during employee interviews. A well-designed PIP by HR can significantly reduce the likelihood of employers losing lawsuits. To that end, it’s recommended that HR professionals thoroughly understand each employee’s case before PIP interviews and remain attentive to non-verbal cues during these discussions. Furthermore, the study found that AI tools leveraging Retrieval-Augmented Generation (RAG) can effectively enhance court judgment classification. Among the tools, Claude 3.7 achieved classification accuracies of 92% and 83%, outperforming ChatGPT-4o’s 87% and 61%. This demonstrates the effectiveness of closed-source generative AI in rapidly categorizing PIP-related judgments. Consequently, it is advised that HR professionals learn AI prompting techniques to productively organize judicial decisions when faced with labor-management disputes.
參考文獻 一、中文 (一)期刊 1.李婉維（2023）。績效改善計畫應用於解僱最後手段性原則之研究。全國律師，27(5)，60-82。 2.林更盛（1998）。論作為解雇事由之「勞工確不能勝任工作」。中原財經法學，4，93-111。 3.張義德（2018）。對於績效考核之司法審查。月旦會計實務研究，12，43-49。 4.陳建文（2009）。業績未達標準作為終止事由之合法性判斷問題／高雄地院九七勞訴三○。臺灣法學雜誌，128，247-250。 5.傅柏翔、王惠玲（2010）。企業績效評估制度對勞動權益之衝擊研究。政大勞動學報，26(12)，91-146。 6.黃程貫（1990）。企業懲罰權。台灣社會研究季刊，2(3&4)，9-73。https://doi.org/10.29816/TARQSS.199012.0001。 7.焦興鎧（1988）。美國法上不當解雇之概念及其救濟之道。美國研究，18(2)，35-131。https://doi.org/10.7015/AS.198806.0035。 (二)書籍 1.三津村直貴、温政堯（2023）。圖解AI人工智慧 = Artificial intelligence。碁峰資訊。97。 2.上野千鶴子著（2021）。如何做好研究論文：成為知識生產者，從提問到輸出的18個步驟(涂紋凰譯)。三采文化。173-174。 3.柯克、陳葵懋、Ryan Chung（2024）。極速ChatGPT開發者兵器指南：跨界整合Prompt Flow、LangChain與Semantic Kernel框架。博碩。2-5與2-6。 4.張永健（2019）。法實證研究：原理、方法與應用。新學林。9與36。 5.陳弘儒（2020）。初探目的解釋在法律人工智慧系統之運用可能。載於李建良(編)，法律思維與制度的智慧轉型。元照。225-299。 6.黃劍青（1985）。勞動基準法詳解。三民。154。 7.蔡炎龍、林澤佑、黃瑜萍、焉然（2024）。少年Py的大冒險：成為Python AI深度學習達人的第一門課(修訂版)。全華圖書。1-32。 (三)政府出版品 1.行政院主計總處（2025年5月），2025年3月底工業及服務業受僱員工人數之表1工業及服務業受僱員工薪資調查統計指標，https://www.dgbas.gov.tw/News_Content.aspx?n=3602&s=234880。 2.國家發展委員會（2024年），中華民國人口推估(2024年至2070年)， https://pop-proj.ndc.gov.tw。 3.勞動部勞動力發展署，就業市場情勢分析2023年11月至2025年2月月報，https://www.wda.gov.tw/News.aspx?n=33&sms=10307&_CSN=25。 (四)碩士論文 1.林武順（1984）。勞工法上解僱問題之研究究〔未出版之碩士論文〕。國立政治大學法律研究所。 2.劉育承（2021）。勞動基準法第十一條第五款解僱事由之研究──以「不能勝任工作」之判斷標準為中心究〔未出版之碩士論文〕。國立臺灣大學法律學系。 (五)雜誌新聞 1.王振容、李瑋（2004年05月25日）。追求績效的人才管理。經濟日報，副刊企管。 2.管婺媛(2024年06月27日)。企業缺工必學、賺錢力高4成秘密，蘋果、台積都在拚DEI。商業周刊，1911期。取自：商業周刊知識庫。 3.蔡茹涵(2023年04月06日)。離職、出缺、找人累壞了？金色三麥讓老鳥全職陪伴菜鳥，保住新人黃金3個月。商業周刊，1847期。取自：商業周刊知識庫。 4.韓化宇(2023年06月08日)。抗缺工》如何讓熟齡夥伴更快上手？養「肌肉記憶」、別一次塞太多東西。商業周刊，1856期。取自：商業周刊知識庫。 二、英文 (一)期刊 1.Alawwad, H. A., Alhothali, A., Naseem, U., Alkhathlan, A., & Jamal, A. (2025). Enhancing textual textbook question answering with large language models and retrieval augmented generation. Pattern Recognition, 162, 111332-. https://doi.org/10.1016/j.patcog.2024.111332. 2.Bailey, M. K., Crolla, L. J., & Sturtevant, A. (2007). Management Q&A. Medical Laboratory Observer, 39(9), 40-41. 3.Beecher, D. D. (2003). The Next Wave of Civil Service Reform. Public Personnel Management, 32(4), 457-474. https://doi.org/10.1177/009102600303200401. 4.Bergman, S., Ji, Z., Kermarrec, A.-M., Petrescu, D., Pires, R., Randl, M., & de Vos, M. (2025). Leveraging Approximate Caching for Faster Retrieval-Augmented Generation. https://doi.org/10.48550/arxiv.2503.05530. 5.Cho, K., Park, Y., Kim, J., Kim, B., & Jeong, D. (2025). Conversational AI forensics: A case study on ChatGPT, Gemini, Copilot, and Claude. Forensic Science International. Digital Investigation (Online), 52, 301855-301865. https://doi.org/10.1016/j.fsidi.2024.301855. 6.Costakis, H. R., & Pickern, J. S. (2022). Managing Human Capital Through the Use of Performance Improvement Plans. The Journal of Applied Business and Economics, 24(6), 216-222. https://doi.org/10.33423/jabe.v24i6.5749. 7.Daley, D. M. (2008). The Burden of Dealing with Poor Performers: Wear and Tear on Supervisory Organizational Engagement. Review of Public Personnel Administration, 28(1), 44-59. https://doi.org/10.1177/0734371X07311253. 8.David, E. M. (2013). Examining the Role of Narrative Performance Appraisal Comments on Performance. Human Performance, 26(5), 430-450. https://doi.org/10.1080/08959285.2013.836197. 9.Falcone, P. (2019). Don’t Sugarcoat Performance Reviews. HRNews. 10.Friel, B. (2005). Red Tape or Redemption. Government Executive, 37(12), 88-88. 11.Goncharsky, A., & Iverson, A. (2018). Human Resource Hot Topics for Closely Held Businesses. Journal of Pension Benefits, 25(3), 22-24. 12.Hawkins, K. (2021). How to Deal with Underperforming Employees. The Lane Report, 36(9), 56-56. 13.Janove, J. (2018). Putting Humanity into HR Compliance: Down with Documentation. HRNews. 14.Janove, J. (2019). Putting Humanity into HR Compliance: Fire Progressive Discipline. HRNews. 15.Janove, J. (2022).How to Maximize the Same Day Summary. HRNews. 16.Jones, M. E. (2018). No. 3 Think Before You Fire. The Army Lawyer, 40-45. 17.Kasson, E. G. (2016). Plan your exits. In HR Magazine ,61(5),110-117. 18.Ladika, S. (2022). Termination Tips for HR Practitioners. HRMagazine, 1-1. 19.Lapi, N. (2018). Performance management in the Vanuatu public service: foundations, achievements and challenges. Asia Pacific Journal of Public Administration = Ya Tai Gong Gong Xing Zheng Xue, 40(4), 245-251. https://doi.org/10.1080/23276665.2018.1543096. 20.Lee, H.-W., & Rhee, D.-Y. (2020). The practices of performance management and low performers in the US Federal Government. International Journal of Manpower, 41(4), 417-433. https://doi.org/10.1108/IJM-12-2018-0404. 21.Leonard, E. (2020). Career Conversations: Progressive Discipline the Right Way. Reference and User Services Quarterly, 59(2), 92–95. https://doi.org/10.5860/rusq.59.2.7272. 22.Martin, J. (2021). It’s time to change how we think about our employees. Strategic HR Review, 20(2), 55–59. https://doi.org/10.1108/SHR-01-2021-0002. 23.McCrea, B. (2021). How to prepare for personnel issues. Independent Banker, 71(5), 22–24. 24.Nagele, L. (2018). 12 Tips for Handling Employee Terminations and Disciplinary Actions. HRNews. 25.Panchal, D., Gole, A., Narute, V., & Joshi, R. (2025). LawPal : A Retrieval Augmented Generation Based System for Enhanced Legal Accessibility in India. https://doi.org/10.48550/arxiv.2502.16573. 26.Rosemarie Lally, J.D. (2021). Agency Must Justify PIP When Employee Challenges Removal. HRNews. 27.Russell, K. (2011). Addressing work performance at an individual level. Strategic HR Review, 10(2), 51-53. https://doi.org/10.1108/shr.2011.37210bab.007. 28.Ryan Scheinkman, Ricardo Cooke, Alexia Vignau, Daniel Green, Philippe Jean-Pierre, & Keyvan Nouri. (2025). Comparing Treatment Recommendations for Ten Dermatological Conditions Using ChatGPT, Claude, and PI AI Models. International Journal of Medical Students, 12-12. 29.Sahoo, C. K., & Mishra, S. (2012). Performance management benefits organizations and their employees. Human Resource Management International Digest, 20(6), 3-5. https://doi.org/10.1108/09670731211260771. 30.Simoneaux, S. L., & Stroud, C. L. (2012). BUSINESS BEST PRACTICES: Great Expectations: Performance Management and Development Strategies. Journal of Pension Benefits, 19(2), 74-76. 31.Smith, J. R. (2020). Performance improvement plan not constructive dismissal. Canadian Employment Law Today, 6-7. 32.Stephens, S. (2018). How to properly fire a practice employee. Medical Economics, 95(8), 31-32. 33.Tolan, T. (2015). Consequences of Performance Improvement Plans. Healthcare Informatics, 32(6), 48-48. 34.Wangsa, K., Karim, S., Gide, E., & Elkhodr, M. (2024). A Systematic Review and Comprehensive Analysis of Pioneering AI Chatbot Models from Education to Healthcare: ChatGPT, Bard, Llama, Ernie and Grok. Future Internet, 16(7), 219-242. https://doi.org/10.3390/fi16070219. 35.West, S. (2016). RE-BALANCING THE PENDULUM: A RECOMMENDATION FOR CIVIL SERVICE REFORM. Administrative Law Review, 68(2), 359–394. 36.Yan, Y., Wang, K., Feng, B., Yao, J., Jiang, T., Jin, Z., Zheng, Y., Zhou, Y., Chen, C., Sui, L., Chen, X., Du, Y., Yang, J., Pan, Q., Zhou, L., Wang, V. Y., Liang, P., & Xu, D. (2025). The use of large language models in detecting Chinese ultrasound report errors. NPJ Digital Medicine, 8(1), 66-79. https://doi.org/10.1038/s41746-025-01468-7. 37.Georgia Institute of Technology, PERFORMANCE IMPROVEMENT PLAN GUIDELINES/PROCESSES, GEORGIA INSTITUTE OF TECHNOLOGY, https://ohr.gatech.edu/sites/default/files/documents/pip_guidelines_process.pdf. 38.Tschohl, J. (2016). When You Have To Say “YOU’RE FIRED.” American Fastener Journal, 32(5), 82–83. (二)書籍 1.Kirkpatrick, D. L. (2006). The Performance Improvement Plan. In Improving Employee Performance Through Appraisal and Coaching,69-79.AMACOM. 2.Nair, V. K., Lavanya, B., Biju, A., Alareeni, B., & Hamdan, A. (2024). Generative AI in Cosmetics Regulations: A Comparison Between ChatGPT, Bard, and Claude. In Navigating the Technological Tide: The Evolution and Challenges of Business Model Innovation ,1081, 82-91. Springer. https://doi.org/10.1007/978-3-031-67437-29. (三)政府出版品 1.美國法規。《美國法典》第43章與第75章，https://www.govregs.com/uscode/title5_partIII。 2.美國政府出版品(1889)。績效管理與獎勵制度重新授權法。https://www.govinfo.gov/content/pkg/STATUTE-103/pdf/STATUTE-103-Pg670.pdf. (四)雜誌 1.Alaniz, R. (2021). Problem Employees – Develop or Dismiss? In Roofing contractor. BNP Media. 14-14. 2.Prencipe, L. (1997). When firing, become a documentarian. In Network world,14(18),93-93. 3.Zavitz, P., & Means, R. (2014). Why Progressive Discipline Systems Often Fail. In Law and order, 62(10),16-20.
描述 碩士
國立政治大學
勞工研究所
108262006
資料來源 http://thesis.lib.nccu.edu.tw/record/#G0108262006
資料類型 thesis
dc.contributor.advisor 劉梅君zh_TW
dc.contributor.advisor Liu, Mei-Chunen_US
dc.contributor.author (Authors) 陳詩吟zh_TW
dc.contributor.author (Authors) Chen, Shih-Yinen_US
dc.creator (作者) 陳詩吟zh_TW
dc.creator (作者) Chen, Shih-Yinen_US
dc.date (日期) 2025en_US
dc.date.accessioned 4-Aug-2025 15:42:17 (UTC+8)-
dc.date.available 4-Aug-2025 15:42:17 (UTC+8)-
dc.date.issued (上傳時間) 4-Aug-2025 15:42:17 (UTC+8)-
dc.identifier (Other Identifiers) G0108262006en_US
dc.identifier.uri (URI) https://nccur.lib.nccu.edu.tw/handle/140.119/158779-
dc.description (描述) 碩士zh_TW
dc.description (描述) 國立政治大學zh_TW
dc.description (描述) 勞工研究所zh_TW
dc.description (描述) 108262006zh_TW
dc.description.abstract (摘要) 　　本文題目「人資與善的距離」，以人資依據解僱最後手段性原則的觀點，思考企業面對不能勝任工作的勞工時，究竟採取多少善意提供勞工改善機會？由於，我國面臨中高齡化的勞動力市場，企業存在勞工高離職率，本研究思考如何留任在職勞工，茲從法理解釋、法院判決與學者論述，觀察前人研究無法改變勞工被資遣，進而關注「績效改善計畫」（簡稱：PIP）輔導不能勝任工作之勞工。企業重視法官的PIP見解，另發現神經網路能將文字轉成數字進行演算，冀望人資不倉促資遣勞工，運用人工智慧（簡稱：AI）分類判決並制定PIP。 　　本研究從收集西文期刊彙整PIP制定過程，另從司法院裁判系統檢索判決全文，關鍵字為PIP、績效改善、績效改進、績效輔導等任何一項，裁判期間設定截至2025年03月31日共498篇PIP判決。其次，採取ChatGPT-4o、Gemini 2.5 Pro、Claude 3.7與Grok 3等四家閉源型生成式AI分類PIP判決，執行描述性統計、分類準確度與卡方獨立性檢定，嘗試找出準確度相對較佳的AI工具；再者，歸納法官肯認企業對不能勝任勞工採取PIP措施且判決資方勝訴的要件。 　　研究發現，西文期刊與我國法院判決在PIP皆運用S.M.A.R.T管理不能勝任勞工，歸納S具體的指「佐證客觀疏失與評估主觀疏失」、M可測量的指「具體指標與量化或順位」、A可達成的指「比較勞工及同事」、R相關的指「達到僱用之客觀合理經濟目的」，T時效的指「改善、追蹤與通知，普遍輔導三個月」；此外，人資在PIP實施漸進式懲戒應落實企業內部辦法，訪談過程採取尊重勞工措施，總之，人資妥善制定PIP能夠降低資方訴訟敗訴，建議人資在PIP訪談前充分瞭解勞工個案且訪談時察言觀色。其次，在AI運用檢索增強生成（簡稱：RAG）有助於分類法院判決，Claude 3.7分類判決準確度為92%與83%，高於ChatGPT-4o準確度為87%與61%，因此，閉源型生成式AI達到快速分類PIP判決，建議人資學習AI提問技巧，遇到勞資糾紛時有能力使用AI整理法官見解。zh_TW
dc.description.abstract (摘要) 　　This study, from a Human Resources (HR) perspective, examines the extent of goodwill and opportunities for improvement that companies should offer to employees deemed incompetent, especially when considering the ultima ratio doctrine of dismissal. Given Taiwan’s aging workforce and high employee turnover rates, this study focuses on strategies for employee retention. While previous legal interpretations, court judgments, and academic literature haven't fully succeeded in preventing terminations, this research highlights the critical role of Performance Improvement Plan(PIP) to guide employees who are incompetent. Recognizing the growing importance of judicial interpretations of PIP for businesses, and the potential of neural networks to process legal texts, the present study advocates for HR professionals to move beyond hasty terminations by leveraging Artificial Intelligence (AI) to classify court judgments and formulate more effective PIP. 　　This study first reviewed the process of designing PIP based on western academic literature. Subsequently, 498 court judgments concerning PIP were gathered from Judicial Yuan’s Judgment System up to March 31, 2025, using keywords such as “PIP”, “performance improvement”, or “performance coaching”. Furthermore, four closed-source generative AI models (i.e., ChatGPT-4o, Gemini 2.5 Pro, Claude 3.7, and Grok 3) were employed to classify these PIP judgments. Descriptive statistics, classification accuracy analysis, and chi-square tests of independence were then performed to identify the AI tool with relatively superior classification accuracy for legal texts. Finally, the judicial arguments that recognize the adoption of PIP by employers as a significant factor for incompetent were also summarized. 　　The findings indicate that both Western academic literature and Taiwanese court judgments apply the S.M.A.R.T. principle when managing incompetent employees through PIP. S (Specific) involves “substantiating objective misconduct and evaluating subjective deficiencies.” M (Measurable) entails “setting concrete indicators and quantifying performance, or ranking.” A (Attainable) means “comparing the employee’s performance with that of colleagues.” R (Relevant) represents “achieving the objective and reasonable economic purpose of employment.” Lastly, T (Timely) denotes “improvement, monitoring, and notification, typically involving a three-month coaching period.” Beyond the S.M.A.R.T. framework, HR professionals should implement PIP in alignment with the principle of progressive discipline and ensure the enforcement of internal company policies, while adopting a respectful approach during employee interviews. A well-designed PIP by HR can significantly reduce the likelihood of employers losing lawsuits. To that end, it’s recommended that HR professionals thoroughly understand each employee’s case before PIP interviews and remain attentive to non-verbal cues during these discussions. Furthermore, the study found that AI tools leveraging Retrieval-Augmented Generation (RAG) can effectively enhance court judgment classification. Among the tools, Claude 3.7 achieved classification accuracies of 92% and 83%, outperforming ChatGPT-4o’s 87% and 61%. This demonstrates the effectiveness of closed-source generative AI in rapidly categorizing PIP-related judgments. Consequently, it is advised that HR professionals learn AI prompting techniques to productively organize judicial decisions when faced with labor-management disputes.en_US
dc.description.tableofcontents 第壹章　緒論 1 　第一節　研究背景 1 　　　　一、中高齡化的勞動力市場 1 　　　　二、高異動率下的留任重要性 2 　　　　三、輔導不能勝任勞工以降低資遣人數 3 　第二節　研究動機 4 　　　　一、法理解釋未修訂「工作確不能勝任」定義 4 　　　　二、學者與判決詮釋「不能勝任」之主客觀論述的實務挑戰 5 　　　　三、學界論述「績效改善」對人資從業者缺乏周全規劃 6 　第三節　研究目的、研究問題及章節 7 　　　　一、從西文期刊彙整PIP制定過程 7 　　　　二、將PIP判決進行實證研究之量化分析 8 　　　　三、以實證研究歸納判決肯認PIP作法 9 　第四節　研究限制 10 　　　　一、研究方法考量個案被認出將影響工作故未採取深度訪談 10 　　　　二、在生成式AI指令輸入的關鍵字影響輸出品質 11 　　　　三、勞工不能勝任判決僅探討PIP個案 11 第貳章　文獻探討 12 　第一節　PIP源於美國聯邦人事制度 12 　　　　一、美國聯邦政府文官改革背景 12 　　　　二、美國聯邦政府關於績效條文 13 　　　　三、美國聯邦政府實施PIP情形 14 　　　　四、績效管理 15 　第二節　績效改善計畫 19 　　　　一、PIP定義 19 　　　　二、PIP制定流程 19 　　　　三、由誰制定PIP？由勞資雙方制定PIP 24 　　　　四、如何制定PIP？從S.M.A.R.T制定PIP 29 　　　　五、何時執行PIP？從警訊判斷時機 32 　　　　六、小結 36 　第三節　PIP落實「漸進式懲戒 37 　　　　一、PIP為解僱最後手段性原則 37 　　　　二、漸進式懲戒定義與PIP關係 38 　　　　三、漸進式懲戒流程 39 　　　　四、未達PIP懲戒措施 41 　　　　五、漸進式懲戒面臨挑戰 43 　　　　六、小結 45 　第四節　人資在PIP協助主管訪談勞工 46 　　　　一、主管畏懼輔導勞工 46 　　　　二、人資在「訪談前」規劃事項 48 　　　　三、人資在「訪談中」應採取的態度 51 　　　　四、人資在「訪談中」妥善記錄 53 　　　　五、人資在「訪談後」善盡職責 59 　　　　六、小結 61 第參章　研究方法與工具 62 　第一節　研究假設與架構 62 　　　　一、研究假設 62 　　　　二、研究架構 62 　第二節　神經網路之人工智慧、機器學習與深度學習 63 　　　　一、神經網路結構為輸入層、隱藏層與輸出層 63 　　　　二、從輸入層辨別人工智慧、機器學習與深度學習 64 　　　　三、深度學習具有三種類型 66 　　　　四、Transformer自注意力改善RNN遞迴計算 68 　第三節　閉源型生成式AI執行量化分析 69 　　　　一、生成式AI沿用Transformer架構 69 　　　　二、檢索增強生成提高生成式AI輸出正確性 70 　　　　三、閉源型生成式AI選用ChatGPT、Gemini、Claude與Grok 71 　　　　四、計算判決分類準確度 75 　　　　五、卡方獨立性檢定 77 　第四節　研究範圍與研究步驟 78 　　　　一、從法院判決篩選關鍵字509篇 78 　　　　二、資料前處理將每篇判決篩出法官心證並存成word檔案 79 　　　　三、定義並標記PIP評價包含加分、扣分與無關 80 　　　　四、定義並標記PIP勝訴包括勞方、資方及無法判斷 81 　　　　五、RAG預訓練PIP加分與PIP扣分 81 第肆章　實證研究結果 82 　第一節　描述性統計 82 　　　　一、ChatGPT、Gemini、Claude與Grok分類「PIP評價」 82 　　　　二、ChatGPT、Gemini、Claude與Grok分類「判決勝訴」 82 　　　　三、以雷達圖分析PIP評價與判決勝訴 83 　第二節　AI分類判決準確度 84 　　　　一、AI分類「PIP評價」之準確度 84 　　　　二、AI分類「判決勝訴」之準確度 84 　　　　三、以熱力圖呈現PIP評價準確度 85 　第三節　卡方獨立性檢定 86 　　　　一、PIP評價對於判決勝訴 86 　　　　二、法院層級對於PIP評價 87 　　　　三、法院層級對於判決勝訴 88 　第四節　從法官見解瞭解PIP趨勢 89 　　　　一、S.M.A.R.T.之S具體的指「提出客觀與主觀之疏失」 89 　　　　二、S.M.A.R.T.之M可測量的指「具體指標與數字或順位」 91 　　　　三、S.M.A.R.T.之A可達成的指「比較勞工及其同事」 92 　　　　四、S.M.A.R.T.之R相關的指「達到僱用之客觀合理經濟目的」 93 　　　　五、S.M.A.R.T.之T時效的指「改善與追蹤」 94 　　　　六、企業執行「漸進式懲戒」按既定辦法且提供多項改善機會 95 　　　　七、PIP寓有最後手段性且趨勢傾向「尊重勞工」 97 　　　　八、小結 98 第伍章　結論與建議 101 　第一節　研究結論 101 　　　　一、從PIP判決制定S.M.A.R.T、漸進式懲戒及尊重勞工措施 101 　　　　二、AI工具有助於快速分類PIP判決 103 　第二節　研究建議 106 　　　　一、人資在PIP訪談前充分瞭解個案並察言觀色 106 　　　　二、人資學習在AI提問並於勞資糾紛時使用AI整理法官見解 107 參考文獻 110 附錄A　法院判決對於「不能勝任工作」之涵攝 118 附錄B　研究步驟 119 附錄C　本文與AI分類「PIP評價及判決勝訴」﹣最高法院 123 附錄D　本文與AI分類「PIP評價及判決勝訴」﹣臺灣高等法院 125 附錄E　本文與AI分類「PIP評價及判決勝訴」﹣臺灣高等法院臺中分院 129 附錄F　本文與AI分類「PIP評價及判決勝訴」﹣臺灣高等法院臺南分院 129 附錄G　本文與AI分類「PIP評價及判決勝訴」﹣臺灣高等法院高雄分院 130 附錄H　本文與AI分類「PIP評價及判決勝訴」﹣臺灣臺北地方法院 131 附錄I　本文與AI分類「PIP評價及判決勝訴」﹣臺灣士林地方法院 135 附錄J　本文與AI分類「PIP評價及判決勝訴」﹣臺灣新北地方法院 136 附錄K　本文與AI分類「PIP評價及判決勝訴」﹣臺灣基隆地方法院 136 附錄L　本文與AI分類「PIP評價及判決勝訴」﹣臺灣桃園地方法院 137 附錄M　本文與AI分類「PIP評價及判決勝訴」﹣臺灣新竹地方法院 138 附錄N　本文與AI分類「PIP評價及判決勝訴」﹣臺灣苗栗地方法院 139 附錄O　本文與AI分類「PIP評價及判決勝訴」﹣臺灣臺中地方法院 140 附錄P　本文與AI分類「PIP評價及判決勝訴」﹣臺灣彰化地方法院 141 附錄Q　本文與AI分類「PIP評價及判決勝訴」﹣臺灣雲林地方法院 141 附錄R　本文與AI分類「PIP評價及判決勝訴」﹣臺灣嘉義地方法院 141 附錄S　本文與AI分類「PIP評價及判決勝訴」﹣臺灣臺南地方法院 142 附錄T　本文與AI分類「PIP評價及判決勝訴」﹣臺灣橋頭地方法院 142 附錄U　本文與AI分類「PIP評價及判決勝訴」﹣臺灣高雄地方法院 143 附錄V　本文與AI分類「PIP評價及判決勝訴」﹣臺灣花蓮地方法院 143zh_TW
dc.format.extent 12616725 bytes-
dc.format.mimetype application/pdf-
dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0108262006en_US
dc.subject (關鍵詞) 績效改善計畫zh_TW
dc.subject (關鍵詞) PIPzh_TW
dc.subject (關鍵詞) S.M.A.R.T.zh_TW
dc.subject (關鍵詞) 漸進式懲戒zh_TW
dc.subject (關鍵詞) 閉源型生成式AIzh_TW
dc.subject (關鍵詞) performance improvement planen_US
dc.subject (關鍵詞) PIPen_US
dc.subject (關鍵詞) S.M.A.R.T.en_US
dc.subject (關鍵詞) progressive disciplineen_US
dc.subject (關鍵詞) closed-source generative AIen_US
dc.title (題名) 人資與善的距離？勞工績效改善計畫之研究 - 以閉源型生成式AI驗證我國法院判決zh_TW
dc.title (題名) How Far is HR from "Goodness"? A Study of Performance Improvement Plans for Incompetent Employees Using Closed - Source Generative AI to Analyze Taiwanese Court Judgmentsen_US
dc.type (資料類型) thesisen_US
dc.relation.reference (參考文獻) 一、中文 (一)期刊 1.李婉維（2023）。績效改善計畫應用於解僱最後手段性原則之研究。全國律師，27(5)，60-82。 2.林更盛（1998）。論作為解雇事由之「勞工確不能勝任工作」。中原財經法學，4，93-111。 3.張義德（2018）。對於績效考核之司法審查。月旦會計實務研究，12，43-49。 4.陳建文（2009）。業績未達標準作為終止事由之合法性判斷問題／高雄地院九七勞訴三○。臺灣法學雜誌，128，247-250。 5.傅柏翔、王惠玲（2010）。企業績效評估制度對勞動權益之衝擊研究。政大勞動學報，26(12)，91-146。 6.黃程貫（1990）。企業懲罰權。台灣社會研究季刊，2(3&4)，9-73。https://doi.org/10.29816/TARQSS.199012.0001。 7.焦興鎧（1988）。美國法上不當解雇之概念及其救濟之道。美國研究，18(2)，35-131。https://doi.org/10.7015/AS.198806.0035。 (二)書籍 1.三津村直貴、温政堯（2023）。圖解AI人工智慧 = Artificial intelligence。碁峰資訊。97。 2.上野千鶴子著（2021）。如何做好研究論文：成為知識生產者，從提問到輸出的18個步驟(涂紋凰譯)。三采文化。173-174。 3.柯克、陳葵懋、Ryan Chung（2024）。極速ChatGPT開發者兵器指南：跨界整合Prompt Flow、LangChain與Semantic Kernel框架。博碩。2-5與2-6。 4.張永健（2019）。法實證研究：原理、方法與應用。新學林。9與36。 5.陳弘儒（2020）。初探目的解釋在法律人工智慧系統之運用可能。載於李建良(編)，法律思維與制度的智慧轉型。元照。225-299。 6.黃劍青（1985）。勞動基準法詳解。三民。154。 7.蔡炎龍、林澤佑、黃瑜萍、焉然（2024）。少年Py的大冒險：成為Python AI深度學習達人的第一門課(修訂版)。全華圖書。1-32。 (三)政府出版品 1.行政院主計總處（2025年5月），2025年3月底工業及服務業受僱員工人數之表1工業及服務業受僱員工薪資調查統計指標，https://www.dgbas.gov.tw/News_Content.aspx?n=3602&s=234880。 2.國家發展委員會（2024年），中華民國人口推估(2024年至2070年)， https://pop-proj.ndc.gov.tw。 3.勞動部勞動力發展署，就業市場情勢分析2023年11月至2025年2月月報，https://www.wda.gov.tw/News.aspx?n=33&sms=10307&_CSN=25。 (四)碩士論文 1.林武順（1984）。勞工法上解僱問題之研究究〔未出版之碩士論文〕。國立政治大學法律研究所。 2.劉育承（2021）。勞動基準法第十一條第五款解僱事由之研究──以「不能勝任工作」之判斷標準為中心究〔未出版之碩士論文〕。國立臺灣大學法律學系。 (五)雜誌新聞 1.王振容、李瑋（2004年05月25日）。追求績效的人才管理。經濟日報，副刊企管。 2.管婺媛(2024年06月27日)。企業缺工必學、賺錢力高4成秘密，蘋果、台積都在拚DEI。商業周刊，1911期。取自：商業周刊知識庫。 3.蔡茹涵(2023年04月06日)。離職、出缺、找人累壞了？金色三麥讓老鳥全職陪伴菜鳥，保住新人黃金3個月。商業周刊，1847期。取自：商業周刊知識庫。 4.韓化宇(2023年06月08日)。抗缺工》如何讓熟齡夥伴更快上手？養「肌肉記憶」、別一次塞太多東西。商業周刊，1856期。取自：商業周刊知識庫。 二、英文 (一)期刊 1.Alawwad, H. A., Alhothali, A., Naseem, U., Alkhathlan, A., & Jamal, A. (2025). Enhancing textual textbook question answering with large language models and retrieval augmented generation. Pattern Recognition, 162, 111332-. https://doi.org/10.1016/j.patcog.2024.111332. 2.Bailey, M. K., Crolla, L. J., & Sturtevant, A. (2007). Management Q&A. Medical Laboratory Observer, 39(9), 40-41. 3.Beecher, D. D. (2003). The Next Wave of Civil Service Reform. Public Personnel Management, 32(4), 457-474. https://doi.org/10.1177/009102600303200401. 4.Bergman, S., Ji, Z., Kermarrec, A.-M., Petrescu, D., Pires, R., Randl, M., & de Vos, M. (2025). Leveraging Approximate Caching for Faster Retrieval-Augmented Generation. https://doi.org/10.48550/arxiv.2503.05530. 5.Cho, K., Park, Y., Kim, J., Kim, B., & Jeong, D. (2025). Conversational AI forensics: A case study on ChatGPT, Gemini, Copilot, and Claude. Forensic Science International. Digital Investigation (Online), 52, 301855-301865. https://doi.org/10.1016/j.fsidi.2024.301855. 6.Costakis, H. R., & Pickern, J. S. (2022). Managing Human Capital Through the Use of Performance Improvement Plans. The Journal of Applied Business and Economics, 24(6), 216-222. https://doi.org/10.33423/jabe.v24i6.5749. 7.Daley, D. M. (2008). The Burden of Dealing with Poor Performers: Wear and Tear on Supervisory Organizational Engagement. Review of Public Personnel Administration, 28(1), 44-59. https://doi.org/10.1177/0734371X07311253. 8.David, E. M. (2013). Examining the Role of Narrative Performance Appraisal Comments on Performance. Human Performance, 26(5), 430-450. https://doi.org/10.1080/08959285.2013.836197. 9.Falcone, P. (2019). Don’t Sugarcoat Performance Reviews. HRNews. 10.Friel, B. (2005). Red Tape or Redemption. Government Executive, 37(12), 88-88. 11.Goncharsky, A., & Iverson, A. (2018). Human Resource Hot Topics for Closely Held Businesses. Journal of Pension Benefits, 25(3), 22-24. 12.Hawkins, K. (2021). How to Deal with Underperforming Employees. The Lane Report, 36(9), 56-56. 13.Janove, J. (2018). Putting Humanity into HR Compliance: Down with Documentation. HRNews. 14.Janove, J. (2019). Putting Humanity into HR Compliance: Fire Progressive Discipline. HRNews. 15.Janove, J. (2022).How to Maximize the Same Day Summary. HRNews. 16.Jones, M. E. (2018). No. 3 Think Before You Fire. The Army Lawyer, 40-45. 17.Kasson, E. G. (2016). Plan your exits. In HR Magazine ,61(5),110-117. 18.Ladika, S. (2022). Termination Tips for HR Practitioners. HRMagazine, 1-1. 19.Lapi, N. (2018). Performance management in the Vanuatu public service: foundations, achievements and challenges. Asia Pacific Journal of Public Administration = Ya Tai Gong Gong Xing Zheng Xue, 40(4), 245-251. https://doi.org/10.1080/23276665.2018.1543096. 20.Lee, H.-W., & Rhee, D.-Y. (2020). The practices of performance management and low performers in the US Federal Government. International Journal of Manpower, 41(4), 417-433. https://doi.org/10.1108/IJM-12-2018-0404. 21.Leonard, E. (2020). Career Conversations: Progressive Discipline the Right Way. Reference and User Services Quarterly, 59(2), 92–95. https://doi.org/10.5860/rusq.59.2.7272. 22.Martin, J. (2021). It’s time to change how we think about our employees. Strategic HR Review, 20(2), 55–59. https://doi.org/10.1108/SHR-01-2021-0002. 23.McCrea, B. (2021). How to prepare for personnel issues. Independent Banker, 71(5), 22–24. 24.Nagele, L. (2018). 12 Tips for Handling Employee Terminations and Disciplinary Actions. HRNews. 25.Panchal, D., Gole, A., Narute, V., & Joshi, R. (2025). LawPal : A Retrieval Augmented Generation Based System for Enhanced Legal Accessibility in India. https://doi.org/10.48550/arxiv.2502.16573. 26.Rosemarie Lally, J.D. (2021). Agency Must Justify PIP When Employee Challenges Removal. HRNews. 27.Russell, K. (2011). Addressing work performance at an individual level. Strategic HR Review, 10(2), 51-53. https://doi.org/10.1108/shr.2011.37210bab.007. 28.Ryan Scheinkman, Ricardo Cooke, Alexia Vignau, Daniel Green, Philippe Jean-Pierre, & Keyvan Nouri. (2025). Comparing Treatment Recommendations for Ten Dermatological Conditions Using ChatGPT, Claude, and PI AI Models. International Journal of Medical Students, 12-12. 29.Sahoo, C. K., & Mishra, S. (2012). Performance management benefits organizations and their employees. Human Resource Management International Digest, 20(6), 3-5. https://doi.org/10.1108/09670731211260771. 30.Simoneaux, S. L., & Stroud, C. L. (2012). BUSINESS BEST PRACTICES: Great Expectations: Performance Management and Development Strategies. Journal of Pension Benefits, 19(2), 74-76. 31.Smith, J. R. (2020). Performance improvement plan not constructive dismissal. Canadian Employment Law Today, 6-7. 32.Stephens, S. (2018). How to properly fire a practice employee. Medical Economics, 95(8), 31-32. 33.Tolan, T. (2015). Consequences of Performance Improvement Plans. Healthcare Informatics, 32(6), 48-48. 34.Wangsa, K., Karim, S., Gide, E., & Elkhodr, M. (2024). A Systematic Review and Comprehensive Analysis of Pioneering AI Chatbot Models from Education to Healthcare: ChatGPT, Bard, Llama, Ernie and Grok. Future Internet, 16(7), 219-242. https://doi.org/10.3390/fi16070219. 35.West, S. (2016). RE-BALANCING THE PENDULUM: A RECOMMENDATION FOR CIVIL SERVICE REFORM. Administrative Law Review, 68(2), 359–394. 36.Yan, Y., Wang, K., Feng, B., Yao, J., Jiang, T., Jin, Z., Zheng, Y., Zhou, Y., Chen, C., Sui, L., Chen, X., Du, Y., Yang, J., Pan, Q., Zhou, L., Wang, V. Y., Liang, P., & Xu, D. (2025). The use of large language models in detecting Chinese ultrasound report errors. NPJ Digital Medicine, 8(1), 66-79. https://doi.org/10.1038/s41746-025-01468-7. 37.Georgia Institute of Technology, PERFORMANCE IMPROVEMENT PLAN GUIDELINES/PROCESSES, GEORGIA INSTITUTE OF TECHNOLOGY, https://ohr.gatech.edu/sites/default/files/documents/pip_guidelines_process.pdf. 38.Tschohl, J. (2016). When You Have To Say “YOU’RE FIRED.” American Fastener Journal, 32(5), 82–83. (二)書籍 1.Kirkpatrick, D. L. (2006). The Performance Improvement Plan. In Improving Employee Performance Through Appraisal and Coaching,69-79.AMACOM. 2.Nair, V. K., Lavanya, B., Biju, A., Alareeni, B., & Hamdan, A. (2024). Generative AI in Cosmetics Regulations: A Comparison Between ChatGPT, Bard, and Claude. In Navigating the Technological Tide: The Evolution and Challenges of Business Model Innovation ,1081, 82-91. Springer. https://doi.org/10.1007/978-3-031-67437-29. (三)政府出版品 1.美國法規。《美國法典》第43章與第75章，https://www.govregs.com/uscode/title5_partIII。 2.美國政府出版品(1889)。績效管理與獎勵制度重新授權法。https://www.govinfo.gov/content/pkg/STATUTE-103/pdf/STATUTE-103-Pg670.pdf. (四)雜誌 1.Alaniz, R. (2021). Problem Employees – Develop or Dismiss? In Roofing contractor. BNP Media. 14-14. 2.Prencipe, L. (1997). When firing, become a documentarian. In Network world,14(18),93-93. 3.Zavitz, P., & Means, R. (2014). Why Progressive Discipline Systems Often Fail. In Law and order, 62(10),16-20.zh_TW