CHS: Small: Narrative-Based Training Simulations Using Theory of Mind

该项目将创建一个基于模拟的培训框架，支持学习者的探索和回放，并将锻炼心理理论技能，以实现社交技能培训的全部承诺。心理理论(Tom)指的是人类使用关于他人心理过程和状态的信念的能力。为了训练社交技能，基于叙事的模拟迅速发展，允许学习者进行角色扮演社交互动。然而，这些系统的设计往往限制了学习者探索不同行为及其后果的能力。试图支持更具生成性的体验的努力，面临着通过交互实现可选路径的组合爆炸，这对开发人员为所有可选路径创建内容提出了压倒性的挑战。相反，培训系统通常是围绕在特定情况下练习特定行为而设计的，这阻碍了学习使用TOM的更一般技能。本研究试图通过三个方面的贡献来解决这个问题：(1)解决生成性的叙事和角色扮演体验概念化的新模式，(2)促进这些生成性体验的内容创建的新方法，以及(3)将心理训练理论嵌入体验中以实现更好的学习结果的方法。这项研究适用于各种情况下的复杂社交技能培训：学校、社区、军队、警察、国土安全和种族冲突。这项研究从范式的转变开始，将社交技能模拟重新概念化为学习者排练一个角色，而不是扮演一个角色。这一转变将利用斯坦尼斯拉夫斯基的主动分析(AA)，这是一种明确练习心理理论技能的表演排练技术。此外，AA分解成短排练场景可以打破长叙事弧线上的组合爆炸，这种组合爆炸加剧了社会培训系统的内容创造。然后，研究将探索在人群来源的数据上使用行为拟合和机器学习技术，作为半自动开发用于社会培训的多代理模拟的方式。这项研究将定量和定性地评估这种方法的能力：(A)提供支持探索和促进TOM使用的经验，以及(B)支持获取可用于利用自动化方法制作这些经验的来自人群的数据。这个项目在结合心理理论建模、互动环境、表演排练和众包方面的前沿工作方面是独一无二的。这种多学科的合作将有助于开发一种创造互动体验的方法，推动社交技能培训的当前技术水平的界限。对众包的依赖提供了一个额外的好处，即能够通过选择相关的人群来引发特定于文化的行为模式，从而允许特定于文化的和跨文化的培训内容。

This project will create a framework for simulation-based training that supports a learner's exploration and replay, and will exercise theory of mind skills, in order to deliver the full promise of social skills training. The term Theory of Mind (ToM) refers to the human capacity to use beliefs about the mental processes and states of others. In order to train social skills, there has been a rapid growth in narrative-based simulations that allow learners to role-play social interactions. However, the design of these systems often constrains the learner's ability to explore different behaviors and their consequences. Attempts to support more generative experiences face a combinatorial explosion of alternative paths through the interaction, presenting an overwhelming challenge for developers to create content for all the alternatives. Rather, training systems are often designed around exercising specific behaviors in specific situations, hampering the learning of more general skills in using ToM. This research seeks to solve this problem through three contributions: (1) a new model for conceptualizing narrative and role-play experiences that addresses generativity, (2) new methods that facilitate content creation for those generative experiences, and (3) an approach that embeds theory of mind training in the experience to allow for better learning outcomes. This research is applicable to complex social skill training across a range of situations: in schools, communities, the military, police, homeland security, and ethnic conflict. This research begins with a paradigm shift that re-conceptualizes social skills simulation as the learner rehearsing a role instead of performing a role. This shift will exploit Stanislavsky's Active Analysis (AA), a performance rehearsal technique that explicitly exercises Theory of Mind skills. Further, AA's decomposition into short rehearsal scenes can break the combinatorial explosion over long narrative arcs that exacerbates content creation for social training systems. The research will then explore using behavior fitting and machine learning techniques on crowd sourced data as way to semi-automate the development of multi-agent simulations for social training. The research will assess quantitatively and qualitatively the ability of this approach (a) to provide experiences that support exploration and foster ToM use and (b) to support acquiring crowd sourced data that can be used to craft those experiences using automatic methods. This project is unique in combining cutting edge work in modeling theory of mind, interactive environments, performance rehearsal and crowd sourcing. This multidisciplinary collaboration will enable development of a methodology for creating interactive experiences that pushes the boundaries of the current state of the art in social skill training. Reliance on crowd sourcing provides an additional benefit of being able to elicit culturally specific behavior patterns by selecting the relevant crowd, allowing for both culture-specific and cross-cultural training content.