인간과 인공지능의 상호작용에서 촉발되는 공동 창의성 연구

박연숙 한국영상학회 2023 한국영상학회 논문집 Vol.21 No.3

This study suggests the significance of the co-creativity of humans and artificial intelligence as a possibility of continuous co-evolution that mutually affects each other. To define the nature of art, humans have constantly researched and conceptualized its properties. Given that art is closely related to human nature, an art language can be an object enjoyed pan-humanly over time and beyond races and geographical distances. The recent achievement of AI has impressive results with regard to the issue of producing artworks, particularly human nature. The ultimate goal of AI developers is to make the most humanized machines, and this study focuses on human-artificial intelligence interaction and co-creativity in this context. Not focusing on the amount of data processed or the speed between humans and AI, the creativity of AI, which occurs through the pastiche, destruction, hybridization or rearrangement of existing information, resembles the structure of human creativity. Co-creativity is established with improvisation in which humans and AI process their respective shares in their respective areas, with these factors jointly executed on a real-time basis. Based on this characteristic, I investigated the possibility of co-creativity between humans and AI in art production. By interacting with AI in real time and responding correspondingly to the deviation actions of AI, humans explore various scripts and coding systems to solve problems in better ways. AI proposes creative clues to interacting humans while approaching the task by reconstructing typical information or providing new perspectives, from which human perception can gain expansion opportunities. Human culture based on human consciousness that expands through AI, rather than attempting to mechanize human capabilities or humanize machines, will be able to reach a higher level. 본 연구는 인간과 인공지능의 공동 창의성의 의의를 상호 영향을 주는 지속적인 공진화의 가능성으로 상정한다. 인간은 예술의 본질을 정의하기 위해 끊임없는 연구와 개념화를 이어왔다. 예술이 인간의 본성과 밀착되었기 때문에, 예술의 언어는 시대와 인종과 지리적 거리감을 초월하여 범인간적으로 향유할 수 있는 대상일 수 있다. 인간 본질의 하나인 예술 창작의 문제를 두고 근래 인공지능의 성취는 인상 깊은 결과를 초래했다. 오늘날 인공지능 개발자의 궁극적 목적은 가장 인간적인 기계로 향해 있는데, 본 연구는 그 한 맥락에서 인간-인공지능의 상호작용과 공동 창의성에 초점을 맞추었다. 인간과 인공지능 사이의 정보 처리량이나 속도에 집중하지 않는다면, 이미 존재하는 정보의 재배치나 해체와 혼성을 통해 발생하는 인공지능의 창의성은 인간 창의성이 발휘되는 구조와 닮아있다. 공동 창의성은 인간과 인공지능이 각자의 영역에서 각자의 몫을 처리하되 실시간 공동으로 실행되는 즉흥성에서 출발한다. 이 특성을 바탕으로 예술 제작에서 인간과 인공지능의 공동 창의성이 발생할 수 있음을 살펴보았다. 인공지능과 실시간 상호작용하고, 인공지능의 편차 반응에 상응하여 대처하면서 인간은 더 새로운 방식으로 문제를 해결할 명령어와 코딩 방식을 연구한다. 또한 인공지능은 인간이 다룰 수 없는 많은 양의 자료를 다루기 때문에 인간이 주목하지 못한 대목에 집중하도록 돕거나 일반적인 경향에 신선한 시각으로 접근한다는 대안을 마련해 줄 수도 있다. 인간에게 창의의 불꽃이 될 수 있는 인공지능의 역할은 전체적 문제를 통제하는 것으로 결말짓는 무엇이 아니라 인간의 혜안을 넓히는 실마리를 제공하는 요소이기에 그 의미가 크다. 인간의 능력을 기계화하거나 기계를 인간화하려는 시도가 아니라 인공지능을 매개로 팽창하는 인간의 의식을 바탕으로 한 인간 문화는 한 차원 높은 단계로 나갈 수 있을 것이다.

대화형 AI 시스템과 비목적성 소통의 사용자 감성경험연구

안성희 한국일러스아트학회 2021 조형미디어학 Vol.24 No.2

In this study, the research of user experience related to conversations with artificial intelligence which is reflected in the recent needs for voice-based interaction was performed. An AI system based on interactions between humans and an AI system that enables conversations such as AI speakers and AI mounted on smartphones was the subject of the experiment. Recently, in addition to the areas that were used in conventional computers or smartphones, such as the transfer of knowledge information and calculations, the emotional area through AI is being attempted with the development of big data deep learning technology. There was a process to find out the emotional connection with AI in the Korean-language conversations, as well as cultural qualities. To verify this, non-purpose conversations were tested and analyzed for users in their twenties by applying an emotional experience research methodology. Voice recognition technologies such as AI speakers are being developed as a self-contained service that focuses too much on online consumer psychological analysis, however, the significance of this study is from the point of view that values and reasons for the existence of technologies, which will be firmly established in people's lives if advanced AI technologies can also develop and install pure functions that help humans. 본 연구는 최근 대두되고 있는 음성기반 인터랙션에 관한 니즈를 반영하여 AI와 사람간의 대화에 관련된 사용자의 경험연구를 진행하였다. 연구에서는 AI 스피커와 스마트폰에 탑재된 AI 등의 구분을 두지 않고 대화가 가능한 AI 시스템과 사람(Human) 간의 인터랙션에 기반한 AI를 실험 대상으로 했다. 또한 한국어를 이용한 대화(Conversation)에서 지식 정보의 전달이나 계산 등 기존의 컴퓨터나 스마트폰에서 문자를 통해서도 해왔던 영역 이외에 빅데이터 딥러닝 기술 등의 발전에 힘입어 최근 시도되고 있는 AI와의 감성 혹은 감정 영역의 대화들을 시도하여 대화(Conversation)가 가지게 되는 인간의 감성적인 연결성 이외에 문화적인 특질들도 나타나는지에 대해서 알아보고자 하였다. 이를 위하여 사용자 감성 경험 연구(EUXR) 방법론을 적용하여 비목적성을 가진 대화를 20대 사용자를 대상으로 실험, 분석 하였다. AI 스피커 등이 지나치게 온라인 소비심리분석에 초점을 맞춘 자기화 서비스로 발전되고 있는 추세 속에서 발전된 AI 기술이 인간 자신의 삶에 도움이 되는 순수한 기능들도 개발되어 탑재할 수 있을 때 기술의 가치와 존재 이유가 더 탄탄하게 사람들의 삶에 자리 잡을 것으로 보인다는 관점에서 본 연구의 의의가 있다고 하겠다.

Utilization of Speculative Design for Designing Human-AI Interactions

장석우,남기영 한국디자인학회 2022 디자인학연구 Vol.35 No.2

Background Understanding artificial intelligence (AI) and contemplating its effects present substantial challenges. A powerful approach to tackling this issue is speculative design (SD), which greatly involves constructing narratives that mobilize discussion on the design and social adoption of technologies. However, research on SD narratives in the context of AI has been scarce. Therefore, this study aims to identify narrative themes in SD that concern human-AI interaction. Methods To begin with, 22 related research cases were collected from the Association for Computing Machinery (ACM) digital library based on selection criteria. Subsequently, a constant comparative method was employed to analyze the selected research cases, which resulted in 16 narratives. Thereafter, affinity diagramming was conducted to form higher-order categories, resulting in identifying five narrative themes. Results The analysis yielded five narrative themes: 1) AI Revealing its Ways of Learning, 2) Exposing the Creator of AI, 3) Staging Conflict among Users, 4) Situating Users as Hackers, and 5) Betrayal of AI. All five narrative themes were found to create a discursive space about human-AI interaction and to generate design insights that concern the socio-technical issues of AI. Conclusions The findings of this study add understanding to the growing field of critical thinking in human-computer interaction (HCI) research. They provide insights into developing more ready- to-use methodological devices that can stimulate discourse around human experience of AI. It is expected that scholars and practitioners alike may use the findings of this study to apply an SD approach for investigating human-AI interaction.

생성형 인공지능 기반 버추얼 휴먼의 언캐니 밸리 효과에 대한 연구

강지영 융복합지식학회 2024 융복합지식학회논문지 Vol.12 No.2

인공지능 기술의 발달로 사람이 해왔던 다양한 분야에 인공지능 버추얼 휴먼의 활용도가 높아지고 있다. 인공지능 휴먼의 발전이 가속화될수록 인간과 인공지능 휴먼의 상호작용에 대한 관심도 커지고 있다. 언캐니 밸리효과(uncanny valley effect)는 일본의 로봇 연구자인 모리 마사히로(Mori Masahiro)가 주창한 이론으로, 인간과 로봇의 상호작용에서 로봇의 움직임이나 생김새가 인간과 더 유사해지면 오히려 불쾌감이 높아지는 현상이다. 그 동안의 많은 관련 연구들은 로봇이나 애니메이션 캐릭터의 이미지와 움직임을 중심으로 연구가 진행되었다. 하지만 본 연구는 생성형 인공지능 기반 버추얼 휴먼의 특성을 살펴보고 이를 통해 발생하는 언캐니 밸리 현상의 존재를 파악하고자 한다. 생성형 인공지능 버추얼 휴먼인 ‘나탈리’를 통해 실험과 설문을 진행하여 생성형 인공지능 버추얼 휴먼의 한계를 보다 정확히 파악하고 향후 더욱 진화된 형태의 인공지능 버추얼 휴먼의 제작에 도움이 되고자 하였다. With the development of artificial intelligence technology, the utility of AI-based virtual humans is increasing across various fields traditionally carried out by humans. As the progress of AI-human development accelerates, there is a growing interest in the interaction between humans and AI entities. The Uncanny Valley Effect, proposed by Japanese roboticist Masahiro Mori, describes the phenomenon where discomfort increases as the appearance and movements of a robot become more human-like. While previous research has predominantly focused on the images and movements of robots or animated characters, this study aims to examine the characteristics of generative AI-based virtual humans. Through experiments conducted with the generative AI-based virtual human ‘Natalie,’ the study seeks to understand the existence of the Uncanny Valley Effect arising from these features. The goal is to precisely identify the limitations of generative AI-based virtual humans and contribute to the development of more evolved forms of AI virtual humans in the future.

Users’ Attachment Styles and ChatGPT Interaction: Revealing Insights into User Experiences

I-Tsen Hsieh(셰이천),Chang-Hoon Oh(오창훈) 한국컴퓨터정보학회 2024 韓國컴퓨터情報學會論文誌 Vol.29 No.3

This study explores the relationship between users attachment styles and their interactions with ChatGPT (Chat Generative Pre-trained Transformer), an advanced language model developed by OpenAI. As artificial intelligence (AI) becomes increasingly integrated into everyday life, it is essential to understand how individuals with different attachment styles engage with AI chatbots in order to build a better user experience that meets specific user needs and interacts with users in the most ideal way. Grounded in attachment theory from psychology, we are exploring the influence of attachment style on users interaction with ChatGPT, bridging a significant gap in understanding human-AI interaction. Contrary to expectations, attachment styles did not have a significant impact on ChatGPT usage or reasons for engagement. Regardless of their attachment styles, hesitated to fully trust ChatGPT with critical information, emphasizing the need to address trust issues in AI systems. Additionally, this study uncovers complex patterns of attachment styles, demonstrating their influence on interaction patterns between users and ChatGPT. By focusing on the distinctive dynamics between users and ChatGPT, our aim is to uncover how attachment styles influence these interactions, guiding the development of AI chatbots for personalized user experiences. The introduction of the Perceived Partner Responsiveness Scale serves as a valuable tool to evaluate users perceptions of ChatGPTs role, shedding light on the anthropomorphism of AI. This study contributes to the wider discussion on human-AI relationships, emphasizing the significance of incorporating emotional intelligence into AI systems for a user-centered future.

AI as a Pair Programming Buddy: Exploring Design Considerations for Human-AI Collaboration

김지연,셰이천,오창훈,최준호 한국지식정보기술학회 2023 한국지식정보기술학회 논문지 Vol.18 No.4

This qualitative study delves into the collaboration experiences of programmers with Artificial Intelligence (AI) in programming, with a focus on key aspects like usability, workload, trust, and learning. Programmers were assigned coding tasks using Copilot, an AI tool, and interviewed to gather their feedback. Effective communication emerges as a pivotal factor in ensuring successful human-AI collaboration. The AI system should grasp intentions, offer accurate code suggestions, provide transparent explanations, and handle errors akin to a human partner. To enhance the programmer's experience with AI-assisted pair programming, the study presents a framework that prioritizes improved communication, code suggestions, explanations, and error resolution. This framework promises to yield a more productive and rewarding experience with AI. The research provides valuable user perspectives, deepening our understanding of AI-assisted pair programming dynamics. Moreover, it lays the groundwork for future AI developments, particularly in supporting programmers. Given the continuous advancements in AI, human-AI collaboration holds tremendous potential to boost productivity and innovation in software development. Leveraging AI's capabilities in programming can unlock greater efficiency and creativity, heralding a promising future for AI integration in the field. Through this investigation, we gain insights that pave the way for a more seamless and productive collaboration between programmers and AI systems. This contributes to the advancement of AI applications in programming domains.

대학생 학습자-AI 간의 교육적 관계에 따른 AI의 기대역할: AI유형 선호도의 조절효과

김진희,리나,이상숙 한국콘텐츠학회 2023 한국콘텐츠학회논문지 Vol.23 No.10

This study aims to investigate the moderating effect of AI preference (three AI types: ITS, ELE, and AI-based Dashboard) of college students between expected relationship building with educational AI and the expected role of AI. For an exploratory study, PLS-SEM (Partial Least Square Structural Equation Model) analysis was conducted based on 425 college students' responses. The results revealed that students expected 'Trust' the most for building relationships with educational AI, and 'ITS' showed the most frequent moderating effects among other AI preferences. The results of this study are expected to be used as basic data to present implications for AI development that can form educationally meaningful interactions with students and the role of AI in collaborating with human instructors to support learning in the educational field.

음성기반 Human-AI Interaction에서 인적요인이 주관적 감성에 미치는 영향

신종규(Jonggyu Shin),허인석(Inseok Heo),Ismatullaev Ulugbek Vahobjon U,김상호(Sangho Kim) 대한인간공학회 2020 대한인간공학회 학술대회논문집 Vol.2020 No.10

Objective: To identify human factors and design parameters that affect user emotion arising from interaction with voice-based intelligent systems. Background: Technology that secures human emotional satisfaction is very important. Recently, AI-infused systems can be developed in the direction of meeting human needs through machine(and deep) learning. Therefore, it is necessary to design guidelines for an intelligent system to evolve in a direction that can secure human emotional satisfaction. Method: Determine human factors and design variables based on prior research. Human factors and design variables are evaluated through experiments using emotional engineering techniques. Emotional factors are derived through factor analysis and emotional differences are analyzed using ANOVA. Results: Among the three human factors (gender, interest in new technology, personality) and the four design speed variables (speech, pitch, structure, and), personality and speed did not affect emotional change. Conclusion: A customized intelligent system can be designed based on human factors and design variables that affect the interaction with the AI-Infused system. Application: This framework will be used as a basic research for integrated emotional evaluation using bio-signals.

생성형 AI를 활용한 창작 과정에서 사용자 의도 반영을 고려한 이미지 변형 기능의 인터랙션 디자인 연구

박성연,허정윤 한국인더스트리얼디자인학회 2024 산업디자인학연구 Vol.18 No.1

연구배경 이미지 생성형 AI의 등장은 사용자에게 이미지 창작 활동을 용이하게 해주었으나, 인공지능 특유의 무작위적 특성으로 인해 사용자가 원하는 이미지를 정확히 얻기 위해서는 시행착오를 반복하게 된다. 대부분의 이미지 생성 AI는 사용자 의도를 반영하기 위한 변형 기능을 제공하고 있지만, 여전히 사용자의 다양한 의도를 완전히 반영하기에는 한계가 있다. 본 연구는 이러한 한계를 극복하기 위해 조형요소를 활용한 사용자 의도의 명확한 표현을 가능하게 하는 이미지 변형 기능을 위한 인터랙션 디자인을 제안하고자 한다. 연구방법 이미지 생성형 AI 사용 경험과 기대하는 변형 요소의 조사를 위해 설문조사를 수행하였다. 설문조사를 통해 도출한 조형요소를 활용한 변형 방식을 이해하기 위해 관찰 실험 및 심층 인터뷰를 수행하였다. 이를 바탕으로 이미지 변형 기능에 대한 인터랙션 아이디어를 도출하였다. 연구결과 설문 조사 및 심층 인터뷰를 통해 이미지 생성형 AI를 활용하여 주로 제작하는 이미지가 인물과 풍경임을 확인할 수 있었고, 이미지 생성 과정의 사용자 의도는 원하는 이미지가 명확한 경우와 새로운 아이디어 및 영감을 얻기 위한 탐색의 두 가지 경우로 나눌 수 있었다. 또한 현재의 이미지 변형 기능은 변형 정도가 제한적이며 변형을 위한 사용자의 의도를 반영하기 어려워 사용자가 인식하는 효용성이 낮음을 확인할 수 있었다. 결론 이미지 생성형 AI를 활용하는 사용자의 의도를 원하는 이미지가 명확한 검색과 영감을 얻기 위한 탐색의 두 가지로 나눌 수 있었으며, 이미지 타입에 따라 변형을 위한 조형 요소가 달라지는 것을 확인할 수 있었다. 이러한 사용자의 의도와 변형을 위한 조형 요소를 반영할 수 있도록 인터랙션 컨셉을 시각화하여 제안하였다. 향후 연구에서 구체적인 인터랙션 방식과 인터페이스 디자인을 제안하고자 한다. Background Most Image Generative AI have an image variation function that helps users change to the image they want. However, there is a limit to reflecting the user's various intentions to use. Therefore, this study proposes an interaction design of an image variation function that can easily express user intentions using formative elements. Methods Through the survey, the experience of using Image Generative AI and expected variation elements were identified, and observation experiments and in-depth interviews were conducted to confirm user intentions based on formative elements. Based on this process, the interaction idea of the image variation function was confirmed. Results The formative elements of person and landscape images were derived, and user satisfaction was high when it was varied similar to the user's expectations. It confirmed the user's intention of creating the wanted image or getting ideas and inspiration. The image variation function did not have a large degree of variation and was difficult to reflect the user's intention, confirming that the usefulness perceived by the user was low. Conclusion In the interaction process, the provision of formative elements suitable for the image category should be considered. In the interaction design process, it is necessary to consider two different modes in consideration of user intention. An interaction method should be provided so that the user can determine the direction of variation using the formative elements.

음성기반 지능형 시스템의 사용자 만족도에 영향을 미치는 인터랙션 설계변수들의 정의 및 분류방안

신종규(Jong Gyu Shin),조인권(In Gwon Jo),임완수(Wan Su Lim),김상호(Sang Ho Kim) 대한인간공학회 2020 大韓人間工學會誌 Vol.39 No.1

Objective: The aim of this study is to identify a few critical design parameters for enhancing user"s satisfaction while interacting with the AI-infused intelligent systems through voice user interface (VUI). Background: The interaction between the user and the AI-infused system is called as Human-AI Interaction (HAII) and supposed to have different features with respect to the human-computer interaction (HCI). It is therefore necessary to establish new criteria for designing and evaluating HAII in the point of user"s satisfaction. Method: This study identified 31 user requirements regarding with HAII from previous studies and organized them into 9 secondary and 3 tertiary level user requirement categories. It was investigated and selected 9 design parameters of VUI that might make differences in user"s satisfaction. The priority of each design parameter was calculated using quality function deployment (QFD) technique. Results: The amount of information, error control, and length of answer were found as the top three critical design parameters among others. They accounted for 51% of the total criticality score. It implies the reliability of information that the AI-infused systems provide during interaction is the most important factor for enhancing user"s satisfaction. Conclusion: This study suggested theoretically nine critical interaction parameters and their priority in designing VUI embedded in AI-infused systems. Application: The result of the study can be used to derive various experimental research models and hypothesis in HAII.