Year: 2023
Type: dissertation
Abstract: The educational role of artificial intelligence-based chatbots is expanding. While AI has been used as a tool for teaching and learning, recently it has had an agency like a teacher to support teaching and learning. As AI has educational agency, the possibility of social and emotional interaction between learners and AI is increasing. Learners are required to have the ability to communicate and collaborate with AI. From the perspective of AI, technology that can provide social and emotional responses to learners is required. However, the AI-based chatbots developed to date have not been able to provide learners with human-like social and emotional responses due to the technological limitations of computer vision and natural language processing, which are factors that hinder interaction and continuous learning with AI. \nApplying anthropomorphic cues to educational AI-based chatbots can elicit social and emotional responses from learners and facilitate their learning. Anthropomorphism is giving human-like characteristics to non-human objects. When anthropomorphism cues such as identity, appearance, and conversation are applied to AI, learners subconsciously perceive AI as a human. Learners perceive AI as a social actor and want to react socially and emotionally to AI. In this process, learners participate more cognitively and emotionally in learning with educational AI chatbots. However, in order for anthropomorphic cues to facilitate meaningful learning without increasing the extrinsic cognitive load of learners, the type and level of anthropomorphism must be carefully considered and designed. \nThis study developed and validated design principles for anthropomorphizing educational AI chatbots. The design and development research methodology was applied, and the specific research questions were 1) What are the design principles of anthropomorphizing of educational AI chatbots, 2) Are the design principles for anthropomorphizing of educational AI chatbots internally valid, and 3) What are the effects of the design principles for anthropomorphizing of educational AI chatbots. The effect of the design principles was verified in terms of perceived anthropomorphism, social and emotional responses, learning engagement, learning outcomes, and usability of chatbots. To develop the design principles for anthropomorphizing educational AI chatbots, literature review and case analysis were conducted. Components, design principles, and detailed guidelines for anthropomorphizing educational AI chatbots were developed. Expert reviews were conducted for internal validation of the design principles, and the design principles were revised based on expert opinions. \nFinally, five components, eleven design principles, and twenty-six detailed guidelines were developed for anthropomorphizing educational AI chatbots. The components are Identity design, Appearance design, Conversation content design, Conversation expression design, and Conversation flow design. The design principles are 1) Principle of educational role assignment, 2) Principle of detailed identity design, 3) Principle of face and body design, 4) Principle of preventing user discomfort, 5) Principle of self-expression, 6) Principle of social expression, 7) Principle of recognition and memory expression, 8) Principle of verbal expression, 9) Principle of non-verbal expression, 10) Principle of early conversation flow design, and 11) Principle of mid-to-late conversation flow design, and detailed guidelines were developed for each principle. \nTo verify the effects of the design principles for anthropomorphizing of educational AI chatbots, an AI chatbot that applied the design principles was developed, and expert reviews were conducted to modify the AI chatbot. A preliminary experiment was conducted with 10 learners to supplement the experimental procedures and tools, and a main experiment (pretest-posttest control group design) was conducted with 70 learners. In the experiment, multimodal data (facial expression, psychophysiological responses, chat log, screen recording, survey, interview data) were collected and analyzed to verify the impact of the anthropomorphic design principles on perceived anthropomorphism, social and emotional responses, learning engagement, learning outcomes, and usability of chatbots. \nThe study found that the anthropomorphic design principles were effective in making learner perceive educational AI chatbots like humans, and they were also effective in inducing social presence and positive or negative emotions, which are social and emotional responses. In addition, the anthropomorphic design principles were found to enhance cognitive and emotional engagement in learning, but did not affect learning outcomes. Learners positively perceived learning with an anthropomorphic AI chatbot and perceived it as useful for learning. Based on these research findings, the effects of the anthropomorphic design principles on perceived anthropomorphism, social and emotional responses, learning engagement, learning outcomes, and usability of chatbots were discussed. \nAs implications for the design and development of anthropomorphizing educational AI chatbots, 1) In order to increase the perceived anthropomorphism of educational AI chatbots, it is necessary to imitate the imperfect appearance of humans, 2) The goal of anthropomorphizing educational AI chatbots is to promote meaningful learning through social and emotional interaction between learner and AI, 3) It is important to adjust the level of anthropomorphizing educational AI chatbots according to learner characteristics, 4) Ethical considerations for the design and development of anthropomorphizing educational AI chatbots and proactive learner education are needed, 5) The design principles for anthropomorphizing educational AI chatbots have the value of cost-effectiveness. \nThis study has theoretical significance in that it developes the design principles for anthropomorphizing AI chatbots for social and emotional interaction between learners and educational AI chatbots. And it scientifically verifies the educational effects of anthropomorphism in terms of learning process and outcomes based on multimodal data. This study has practical significance in that it provides specific prescriptions for designing AI whit an agency that interacts socially and emotionally with learners, breaking away from the conventional perspective that treats AI as a tool for teaching and learning. This study is expected to provide a cornerstone for the development of AI for collaboration and coexistence with learners.
Author: 한예진
Language: Korean
Subfield: Social Psychology
Field: Psychology
Domain: Social Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7072724052
