Using Conversational Agents to Foster Preschool Children's Science Learning and Engagement from Interactive Science Videos

科学电视节目是学龄前儿童非正式学习和丰富知识的重要来源。然而，电视节目的一个限制是，它在很大程度上是一种单向的、非互动的媒体。研究表明，儿童通过积极参与内容学习效果最好，父母可以通过与孩子共同观看和交谈，使电视观看更具互动性。然而，许多父母和其他成年人可能缺乏时间或经验，也不熟悉科学语言和内容，无法为他们的孩子提供关键的及时支持。这项研究试图利用人工智能的最新进展，现在可以让孩子们愉快地与自动对话代理进行互动。研究小组将探索这种作为科学视频中的屏幕角色嵌入的对话代理是否可以通过模拟与成年人共同观看的好处，与孩子们就节目的科学内容进行有意义的互动。如果成功，该项目将为一种新的科学节目类型奠定基础，有助于将视频观看转变为更具互动性和吸引力的学习体验。该项目由推进非正式STEM学习（AISL）计划资助，该计划旨在推进非正式环境中STEM学习的设计和开发的新方法和基于证据的理解。这包括提供多种途径，扩大STEM学习经验的获取和参与，推进非正式环境中STEM学习的创新研究和评估，以及发展参与者对深度学习的理解。该项目将在未来的儿童动画电视节目中制作三集11分钟的互动视频，其中包含一个对话代理。这些视频将让孩子们在主角解决日常科学难题的过程中与主角对话，从而通过架起的对话培养孩子们观察、预测、发现模式和解决问题的能力。这项研究将在两个迭代周期中进行，目标是为每一集开发和测试嵌入式会话功能。在每个周期中，由儿童电视制作专家、教育和HCI研究人员组成的项目团队将开发故事板、对话提示和后续跟进，根据修改后的脚本制作动画视频，并创建与对话代理集成的交互式视频移动应用程序。将对10名儿童进行实地测试，以迭代改进嵌入式会话功能。在试点测试阶段，将进行一项对照研究，每组30名儿童（N=120）： 1)观看嵌入会话功能的剧集；2)与真人搭档一起观看剧本中的对话，而不是与虚拟角色一起观看；3)观看带有伪互动的剧集，即动画人物提问，但并不试图理解或亲自回应儿童的回答；4)看没有对白的剧集。从实验中收集的数据将用于检查会话代理的使用是否以及以何种方式影响儿童的参与、注意力、沟通策略、感知和科学学习，以及这些影响是否因儿童的年龄、性别、社会经济地位、语言背景和英语口语熟练程度而异。该项目将全面评估将对话代理纳入屏幕媒体以促进幼儿STEM学习和参与的可行性和潜力。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Science television shows are an important source of informal learning and enrichment for preschool-aged children. However, one limitation of television programming is that it is largely a one-way, non-interactive medium. Research suggests that children learn best through active engagement with content, and that parents can make TV watching more interactive by co-viewing and talking with their children. However, many parents and other adults may lack the time or experience and comfort with science language and content to provide critcial just-in-time support for their children. This study seeks to take advantage of recent advances in artificial intelligence that now allow children to enjoyably interact with automated conversational agents. The research team will explore whether such conversational agents, embedded as an on-screen character in a science video, can meaningfully interact with children about the science content of the show by simulating the benefits of co-viewing with an adult. If successful, the project could lay the foundation for a new genre of science shows, helping transform video watching into more interactive and engaging learning experiences. This project is funded by the Advancing Informal STEM Learning (AISL) program, which seeks to advance new approaches to, and evidence-based understanding of, the design and development of STEM learning in informal environments. This includes providing multiple pathways for broadening access to and engagement in STEM learning experiences, advancing innovative research on and assessment of STEM learning in informal environments, and developing understandings of deeper learning by participants.This project will develop interactive videos incorporating a conversational agent in three 11-minute episodes of a future children's animated television program. The videos will enable children to speak with the main character of the show as the character solves everyday science mysteries, thus priming children to engage in observation, prediction, pattern finding, and problem solving through scaffolded conversation. This study will be carried out in two iterative cycles with the goal of developing and testing the embedded conversational function for each episode. In each cycle, the project team, which includes experts in children's TV production, as well as educational and HCI researchers will develop the storyboard and conversation prompts and follow-ups, create animated videos based on the revised script, and create a mobile application of the interactive video integrated with the conversational agent. Field testing with 10 children will be conducted to iteratively improve the embedded conversational function. In the pilot testing stage, a controlled study will be conducted with 30 children in each group (N=120): 1) watching the episode with the embedded conversational function; 2) watching the episode with a human partner carrying out the dialogue in the script rather than the virtual character; 3) watching the episode with pseudo-interaction, in which the animated character asks questions but does not attempt to understand or personally respond to children's answers; and 4) watching the episode with no dialogue. Data collected from the experiments will be used to examine whether and in what ways use of a conversational agent affects children's engagement, attention, communication strategies, perceptions, and science learning, and whether these effects vary by children's age, gender, socioeconomic status, language background, and oral language proficiency in English. The project will provide a comprehensive evaluation of the feasibility and potential of incorporating conversational agents into screen media to foster young children's STEM learning and engagement.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.