SBIR Phase I: A Novel Human Robot Interaction System Using Affective Theory-of-Mind Computing to Improve User-Responses and Efficacy of Automated Tutoring

这个小企业创新研究(SBIR)第一阶段项目通过开发情感反应系统的基础来促进人工智能(AI)和机器人技术的发展。该项目以心理学和机器人学中的心理理论研究为基础，专注于提高机器人整合两条信息以确定第三条信息的能力。这些推论可以是具体的，也可以是抽象的，就像对另一个代理人--包括人类用户--的信念、愿望和意图的猜测一样抽象。该项目为机器人的人工智能系统创建了一个框架，以维护世界的几个相互竞争的模型(粒子)，并选择和修改那些与现实世界匹配可能性更高的模型(过滤)。这种粒子滤波方法是一种在复杂环境中导航的有效策略。核心研究集中于在抽象环境中适应和评估粒子过滤方法，包括知识、情感和目标状态。如果成功，随着机器人通过传感器和语言处理获得更多信息，它将调整用户情绪状态的模型，并相应地调整其交互。从本质上讲，这一综合框架将使机器人能够理解人类的愿望和挫折，以找到针对这些需求的创造性解决方案。这个项目更广泛的影响/商业潜力是一种人工智能界面的演变，它模拟了一种同理心的人与人之间的体验。由于该系统将代表和跟踪有关人类用户的多个维度，因此机器人可以识别用户感到困惑或沮丧的时候，并做出相应的反应。这个直观、自然主义的人工智能平台拥有大量的研究、教育和治疗应用以及商业化渠道。凭借其开源软件基础设施和该平台的相对低成本，最初的商业目标是作为一个强大的科学、技术、工程和数学(STEM)教育平台，以及作为自闭症儿童和成人的治疗技术。

This Small Business Innovation Research (SBIR) Phase I project advances artificial intelligence (AI) and robotics by developing the foundation for emotionally responsive systems. Building on Theory of Mind research in psychology and robotics, this project focuses on enhancing a robot's capacity to integrate two pieces of information in order to determine a third piece. These inferences can be concrete or as abstract as conjectures about another agent's - including a human user's -beliefs, desires, and intentions. The project creates a framework for the robot's AI system to maintain several competing models of the world (particles) and select and modify those models that have a higher likelihood of matching the real world (filtering). This method of "Particle Filtering" is an effective strategy or navigating complex environments. The core research focuses on adapting and assessing particle filtering methods within abstract contexts including knowledge, emotional, and goal states. If successful, as the robot gains additional information through sensors and language processing, it will adapt its models of the users' emotional state and adjust its interactions accordingly. Essentially, this integrative framework will enable robots to understand human desires and frustrations to find creative solutions in response to those needs. The broader impact/commercial potential of this project is the evolution of an Artificial Intelligence interface that emulates an empathic human-to-human experience. Because the system will represent and track multiple dimensions about a human user, the robot can recognize when a user is confused or frustrated and respond accordingly. This intuitive, naturalistic AI platform has substantial research, education, and therapeutic applications and commercialization channels. With its open source software infrastructure and the relative low-cost of the platform, the initial commercial targets are as a platform for robust Science, Technology, Engineering, and Mathematics (STEM) education and as therapeutic technology for children and adults on the autism spectrum.