DSpace Collection:
https://ah.lib.nccu.edu.tw/handle/140.119/51522
2024-03-28T14:33:43Z虛擬代理人的相互性行為對信任之影響
https://ah.lib.nccu.edu.tw/handle/140.119/147168
題名: 虛擬代理人的相互性行為對信任之影響; Agents’ Mutuality Behavior and Human Trust in Human-Agent Interaction
Authors: 林庭羽; Lin, Ting-Yu
摘要: 近年來,隨著人工智能的進步,電腦逐漸擁有強大的決策能力,這對人類的生活和工作方式都產生了變革性的影響,也讓人與代理人互動(Human-Agent Interaction, HAI)的形式更加多元,AI可化身為能協助人類行事具形體或不具形體的代理人(agent),像是智能機器人、自動駕駛汽車等,這些代理人會協助人們完成任務。而「信任」是人與代理人得以不斷合作與互動的重要關鍵之一,因此我們需要研究是什麼因素能促進或減低人們對AI的信任,尤其是以AI為基礎的代理人之信任。\n 在代理人的設計中,人們通常依據過去與他人互動經驗之心智模型來與代理人互動。根據溝通調適理論,人們常常調整自己的語言、非語言和副語言行為,以適應他人的溝通特徵,減少社會距離。本研究即探討在虛擬環境中當代理人呈現具眼神注意及調節行為時,人們是否會因此對代理人產生信任?為了驗證假設,本實驗為二因子組間設計:代理人的外觀(人形/機器人型)和注意行為(有注意行為/無注意行為)兩個主變項。實驗方式為在虛擬實境環境中進行人與代理人共同協作拼圖任務,由代理人給予受試者拼圖指示和建議,受試者可自行決定是否要接受代理人之指示,並於實驗後填寫綜合信任技術接受模型問卷,了解代理人的外觀擬人化與注意行為變化是否會對人類產生信任影響。\n 實驗結果發現,我們無法藉由操弄來達到原先想形塑的社會知覺(Social Perception),因人們對於代理人的注意行為會產生不同的主觀感知,本實驗也再次驗證人與代理人互動中心智模型的重要性。在實驗中,若代理人行為表現與人相似且具有相應的注意行為,可以符合人們的心智模型時,會增加對代理人的可預測性,而讓人更有安全感,以致可以專注完成任務,所以任務表現較佳。但若實驗情境無法與心智模型相呼應,因降低了對代理人的可預測性也讓任務表現較差,更有可能受到代理人「外觀」與「注意感覺」的干擾影響表現。另外,受試者對於人形外觀的高感知能提升對代理人的誠信感;當受試者認為代理人具備注意行為時,他們更信任代理人並願意接受其指示,且能提升彼此合作關係。儘管本研究與原先假設並不完全符合,但實驗結果出乎我們的想像令我們獲得許多有趣的發現,期望為日後的HAI領域提供新的發展方向。
描述: 碩士; 國立政治大學; 數位內容碩士學位學程; 1094620132023-09-01T07:56:23Z《追蹤一道虛擬的光》——移動投影裝置創作論述
https://ah.lib.nccu.edu.tw/handle/140.119/147167
題名: 《追蹤一道虛擬的光》——移動投影裝置創作論述; Creation of a Movable Projection Installation "A Traced Ray"
Authors: 黃建榮; Huang, Chien-Jung
摘要: 隨著各家科技巨擘提出對於元宇宙的願景,在當代科技推波助瀾之下,人們生活的軌跡逐漸涵蓋虛擬空間,因此本創作研究將主軸聚焦於探討人、現實空間、虛擬空間的關係美學,並且透過實踐開發新型態的影像裝置的過程,進行行動研究,於創作實踐中的產生的問題,藉由實驗與檢討,不斷修正思路,並且透過進行文獻探討,借助理性的思辯方式,深入探究創作行為背後的哲學意涵。爬梳研究相關的藝術典範、科技哲學、社會現象等,以加深對創作的理解和認識。本創作研究透過軟硬體開發整合,以遊戲引擎Unreal Engine 5為開發平台,視覺呈現以即時光線追蹤系統為核心,以手機作為移動感測器,結合微型投影機,打造「移動投影裝置」。並且創作現地製作的新媒體藝術作品,進行展覽現場的3D掃描。透過作品展覽展出的過程,觀察體驗者的反應與回饋。最終,進行回顧與省思,本創作研究為個人未來的創作和研究提供重要的基礎和啟發。; As tech giants revealed their visions for the Metaverse, the relentless advancement of technology has increasingly ushered people`s lives into the domain of virtual spaces. This study places its focus on the exploration of the Relational Aesthetic among individuals, physical spaces, and the virtual spaces. Through practical implementation, navigating an innovative form of the new media installation, the study`s methodological evolution is shaped by action research, which actively involving problem-solving and iterations during the creative process. Thorough analysis of related art paradigms, technological philosophies, and phenomenalism enhances the study`s comprehension.\nThis study deploys integration of software and hardware, leveraging the real-time ray tracing technology in Unreal Engine 5 for visual presentation, while utilizing mobile devices as motion sensors and micro projectors to construct a "Movable Projection Installation." Additionally, it creates new media art pieces on-site, incorporating 3D scanning of the exhibition venue. Via observing audience responses and feedback during the exhibition, the study culminates in a reflective analysis. Ultimately, this creative exploration serves as a significant foundation and inspiration for the future works.
描述: 碩士; 國立政治大學; 數位內容碩士學位學程; 1084620122023-09-01T07:56:11Z遊戲行為設計對於態度極化之影響 —— 以死刑議題為例
https://ah.lib.nccu.edu.tw/handle/140.119/147166
題名: 遊戲行為設計對於態度極化之影響 —— 以死刑議題為例; The Impact of Game Behavior on Attitude Polarization —— Taking the Death Penalty Issues as an Example
Authors: 林修廷; Lin, Xiu-Ting
摘要: 近來的態度極化現象研究多以社群平台上的內容,尤其是各類政治議題為主;然而態度極化的議題絕不止於政治議題,造成極化的傳播方式也不限於社群平台,各領域中的爭議性主題皆可能導致極化的發生。以臺灣社會為例,死刑議題即是一項持續受到社會各方關注與爭辯的議題,支持與反對方立場壁壘分明,明顯地呈現兩極化。\n\n另一方面,近年來的研究指出,遊戲具備的互動性比起靜態的文本,能夠讓參與者有更豐富的感知體驗,進而深刻理解遊戲欲傳達的訊息;和其他媒介相比能帶來更持久的認知變化。因此,本研究便以此為途徑,企圖探究遊戲與態度極化間的相互關係。我們以死刑議題為參考,假設受試者對議題的原始態度(支持/不支持)、其接觸的素材形式(遊戲/文字內容)及其議題立場(支持/不支持)相互作用,希望透過一個 2 × 2 × 2 的組間設計實驗,觀察此三項因素的相互作用是否影響受試者實驗前後的態度產生變化。\n\n在實驗結果之中,我們觀察到受試者的態度變化幅度雖不明顯,但仍可呼應部分的研究假設:遊玩遊戲的受試者確實受到素材形式及其議題立場的影響,其態度的變化趨勢與研究假設大致相符,證實遊戲在其中的確具備一定的影響力。另一方面,受試者對死刑議題的態度並未因實驗的影響而變得更加強烈,事實上也遏止了極化的發生,進而可能降低受試者對於死刑議題的偏見。藉由此研究,我們期望能為日後的相關研究提供不同的思路,以及包含實驗素材設計、問卷量表設計與議題選擇等研究設計面向的建議與參考。; Past research on attitude polarization focused on radicalized political attitudes resulting from interactions over social media platforms, especially various political issues. However, the scope of topics driving attitude polarization extends beyond politics, and the propagation of polarization is not limited to social media. Controversial subjects across different domains have the potential to trigger polarization. Take the example of Taiwanese society where the issue of the death penalty continues to be a subject of ongoing societal concern and debate, with clear-cut positions in favor of and against, vividly demonstrating a dichotomy.\n\nOn the other hand, recent research has indicated that the interactivity inherent in games, compared to static text, allows participants to have richer perceptual experiences, leading to a deeper understanding of the intended message and potentially fostering more enduring cognitive changes. Therefore, this study takes this avenue to investigate the interplay between gaming and attitude polarization. Using the death penalty as a reference point, we hypothesize that the participants` original attitudes towards the issue (for or against the death penalty), the format of material they engage with (game or texts), and the stance on the issue (for or against the death penalty) interact. We propose a 2x2x2 between-subject design experiment, aiming to observe whether the interaction of these three factors influences changes in participants` attitudes before and after the experiment.\n\nIn the results of the experiment, we observed that the extent of participants` attitude changes wasn`t markedly pronounced. However, this still aligns with some of the research hypotheses: participants engaged in gaming did indeed exhibit influences from the format of the material and the stance on the issue. The general trend of their attitude changes roughly corresponds to the research assumptions, confirming that game factor do possess a certain degree of influence within this context. On the other hand, participants` attitudes towards the death penalty did not intensify due to the experimental influence. In fact, it also curbed the occurrence of polarization, potentially mitigating biases concerning the death penalty issue. Through this study, we aspire to provide different avenues for future related research, encompassing recommendations and references for aspects of research design such as experimental material creation, questionnaire scale formulation, and topic selection.
描述: 碩士; 國立政治大學; 數位內容碩士學位學程; 1084620072023-09-01T07:55:57Z結合 AI 虛擬助教與沉浸式學習策略於程式教學遊戲的設計與評估
https://ah.lib.nccu.edu.tw/handle/140.119/147164
題名: 結合 AI 虛擬助教與沉浸式學習策略於程式教學遊戲的設計與評估; Design and Evaluation of a Programming Learning Game Incorporating AI Virtual Assistants and Immersive Learning Strategies
Authors: 古健樺; Gu, Jian-Hua
摘要: 學習程式設計已經是全世界的運動,特別是非資訊背景相關的學生在學習上會遇到非常多的困難,所以本研究設計了《Code Bunny》的 JavaScript 程式語言沉浸式學習遊戲,並加入生成式人工智慧 (ChatGPT) 技術實作之虛擬助教融入鷹架理論,除了提供學習者一個互動性強、沈浸式且具有學習鷹架的學習環境外,結合虛擬助教進行解答,引導學習者透過解決程式設計問題,循序漸進地學習和應用 JavaScript 程式語言。本研究探討加入 AI 虛擬助教的沉浸式學習模式,實施於非資訊背景的大學生與研究生學習 JavaScript 程式設計,欲了解此學習模式對學習者的學習成效、學習動機、心流狀態、活動焦慮、三階段的情緒狀態、綜合學習心得回饋與學習過程的行為模式之影響。研究對象為非資訊背景的大學生與研究生,總共 41 位實驗對象,實驗採用單一前後測設計,在受測者進行學習活動前,需填寫遊戲實驗同意書,並且進行先備知識前測,體驗完學習遊戲後,受測者需要再次填寫程式學習成效後測,此後測題目與前測相同,但是題號順序不同,接著完成 ARCS 動機量表、心流量表、活動焦慮量表和動機及情緒三階段量表,最後填寫一份質性的心得問卷,在遊戲體驗的過程中會要求受測者對螢幕進行錄影,除了分析上述的量表之外,其中實驗後的心得問卷將以質性方法進行分析,實驗也會將螢幕錄影內容編碼後進行序列分析,以了解受測者在使用 AI 虛擬助教輔助程式學習的行為模式。研究結果顯示,使用AI虛擬助教輔助沉浸式學習能夠有效提升學習者的JavaScript學習成效、動機和心流狀態。然而,仍有一些改進的空間,包括提升助教回答的準確性和指導性,改進遊戲介面的清晰度,以及更好地引導和延伸學習的能力。這些建議和改進將有助於進一步提升AI虛擬助教的效果和學習體驗。; Learning programming is already a movement all over the world, especially students with non-information background will encounter a lot of difficulties in learning, so this study designed a JavaScript immersive learning game called "Code Bunny", and incorporating the virtual assistant into the implementation of generative AI (ChatGPT) technology into the scaffolding theory, in addition to providing learners with a highly interactive, immersive environment with learning scaffolding, the virtual teaching assistant is introduced to answer questions and guide learners through learn and apply the JavaScript programming step by step by solving programming problems. This study explores the immersive learning mode of adding AI virtual teaching assistants to learn JavaScript programming. The subjects of the study were 41 undergraduates and postgraduates with non-information background. This study is a single-group quasi-experimental design. The purpose of this experiment is to explore learners` learning effectiveness, learning motivation, flow state, learning anxiety, and emotional state. The research results show that the use of AI virtual teaching assistants to assist immersive learning can effectively improve learners` JavaScript learning effectiveness, motivation and flow state. However, there is still some room for improvement, including improving the accuracy and instructiveness of TA responses, improving the clarity of game interfaces, and the ability to better guide and extend learning. These suggestions and improvements will help to further enhance the effect and learning experience of AI virtual teaching assistants.
描述: 碩士; 國立政治大學; 數位內容碩士學位學程; 1064620112023-09-01T07:55:40Z