大型語言模型表達建議確信度對創業者心理狀態與決策傾向影響之研究;A Study on the Effects of Large Language Models’ Expressed Advice Confidence on Entrepreneurs’ Psychological States and Decision-Making Tendencies

NCU Institutional Repository > 管理學院 > 資訊管理研究所 > 博碩士論文 > Item 987654321/99392

請使用永久網址來引用或連結此文件: https://ir.lib.ncu.edu.tw/handle/987654321/99392

題名:	大型語言模型表達建議確信度對創業者心理狀態與決策傾向影響之研究;A Study on the Effects of Large Language Models’ Expressed Advice Confidence on Entrepreneurs’ Psychological States and Decision-Making Tendencies
作者:	張允澔;Chang, Yun-Hao
貢獻者:	資訊管理學系
關鍵詞:	大型語言模型;AI 信任;創業自我效能;市場驗證意圖;AI 素養;結構方程模型
日期:	2025-12-18
上傳時間:	2026-03-06 18:53:00 (UTC+8)
出版者:	國立中央大學
摘要:	過去研究指出，大型語言模型常以高度自信的語氣提供創業建議，此種過度自信語氣一方面可能提升創業信心，另一方面亦可能強化對 AI 的依賴，削弱對市場驗證的重視。本研究以大型語言模型建議語氣的確信度為自變項，透過情境式實驗結合問卷調查，比較高確信與校準不確定兩種語氣，並採用偏最小平方法結構方程模型分析對 AI 的信任、創業自我效能、創業意向與市場驗證意圖之路徑關係，同時檢驗 AI 素養的調節效果。研究結果顯示：高確信語氣會提高受試者對 AI 的信任與創業自我效能，進而強化投入創業行動的意圖。同時，較高的AI信任會降低市場驗證意圖，顯示在高度信任 AI 建議時，受試者較不積極從事市場資訊蒐集與驗證活動，而創業自我效能對市場驗證意圖的影響則未達顯著水準。進一步分析亦發現，AI 素養雖未明顯改變建議語氣對 AI 的信任之影響，卻會減弱高確信語氣對創業自我效能的提升效果，使創業意向與市場驗證意圖較不完全受單一語氣風格牽動，顯示 AI 素養在面對過度自信的 AI 建議時具有一定的校準與保護功能。;Prior research indicates that large language models (LLMs) often deliver entrepreneurial advice in a highly confident tone. Such overconfident language may, on the one hand, enhance entrepreneurial confidence, but on the other hand, strengthen dependence on AI and weaken the emphasis on market validation. This study takes the confidence level of an LLM’s advisory tone as the independent variable and, through a scenario-based experiment combined with a questionnaire survey, compares a high-confidence tone with a calibrated-uncertainty tone. Partial least squares structural equation modeling (PLS-SEM) is employed to analyze the path relationships among trust in AI, entrepreneurial self-efficacy, entrepreneurial intention, and market validation intention, while also examining the moderating effect of AI literacy. The empirical results show that a high-confidence tone increases participants’ trust in AI and their entrepreneurial self-efficacy, thereby strengthening their intention to engage in entrepreneurial action. At the same time, higher trust in AI reduces market validation intention, suggesting that when AI advice is highly trusted, participants are less inclined to conduct market information gathering and validation activities, whereas the effect of entrepreneurial self-efficacy on market validation intention does not reach statistical significance. Further analysis reveals that although AI literacy does not substantially change the effect of advisory tone on trust in AI, it attenuates the positive impact of a high-confidence tone on entrepreneurial self-efficacy, making entrepreneurial intention and market validation intention less fully driven by a single tone style. This suggests that AI literacy plays a calibrating and protective role when users are confronted with overconfident AI advice.
顯示於類別:	[資訊管理研究所] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	6	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....