EAGER: Toward Automatic Generation of Challenge-Driven Interactive Narratives

这个项目将进行计算系统的第一次探索，该系统共同生成相互交织的叙事和游戏进度。该系统将利用一种名为Ludo-Narrative Units(LNU)的小说表示法，表示叙事、游戏挑战和选择的组合。利用考虑了玩家代理、叙事理解和游戏学习目标等现象的受众体验模型，系统将推理叙事选择、叙事阐述和游戏挑战的进展，并可以动态重塑内容以匹配指定的目标和受众上下文。为了实现这一点，该项目将建立一个人工智能导演，在观众参与基于游戏的叙事期间为他们构建互动场景。从创作的LNU池中动态构建场景已经包含了比目前在最先进的场景驱动游戏和其他流行的互动叙事中出现的更大的灵活性，因为LNU可以根据观众的活动以多个顺序选择。但设想的系统将通过使用故事生成技术修改LNU内容来提供进一步的灵活性和定制化。具体地说，它将使用基于想象回忆的故事规划系统来动态提供这些内容。此外，将通过使用约束求解算法来提供全局叙事和游戏结构，以创建满足各种系统目标的完整体验，使用答案集编程技术编写。这实现了至关重要的灵活性，允许系统对任意数量的域进行推理，而不会将其锁定在预先计划的接口和基础架构中。这项研究的高级影响将有助于更深入地理解这些场景，因为它使用生成所需的严格级别来定义它们。在更技术性的层面上，这项研究将产生第一个在共享体验目标指导下生成混合互动叙事和游戏的系统。从事这项工作的高级动机之一是它对教育具有潜在的深远影响，使下一代教育软件能够受益于场景的激励力量，并专门针对特定学习者的学习概况、目标和进步而生成。以这种方式产生的体验，并基于这些模型，也可能更有效地接触到以游戏为主要媒体形式长大的受众。

This project will conduct the first exploration of a computational system that co-generates intertwined narrative and gameplay progressions. The system will utilize a novel representation called ludo-narrative units (LNUs), representing combinations of narrative, gameplay challenges, and choices. Utilizing audience experience models taking into account phenomena such as player agency, narrative comprehension, and gameplay learning objectives, the system will reason about progressions of narrative choices, narrative exposition, and gameplay challenges, and can dynamically reshape content to match designated goals and audience context. To accomplish this, the project will build an artificial intelligence director to construct interactive scenarios for audiences during their engagement with a game-based narrative.The dynamic construction of scenarios from a pool of authored LNUs would already comprise more flexibility than currently appears in state-of-the-art scenario driven games and other popular interactive narratives, since LNUs can be selected in multiple orders in response to audience activity. But the envisioned system will provide further flexibility and customization through modification of LNU content using story generation technology. Specifically, it will use a story-planning system based on imaginative recall to dynamically provide this content. Further, global narrative and gameplay structure will be provided by using a constraint-solving algorithm to create full experiences satisfying the various system goals, written using answer set programming techniques. This enables the crucial flexibility that allows the system to reason over an arbitrary number of domains without locking it into pre-planned interfaces and infrastructure.A high-level impact of this research will be contributing to a deeper understanding of these scenarios, as it defines them with the level of rigor required for generation. At a more technical level, the research will produce the first system that generates mixed interactive narrative and gameplay guided by shared experience goals. One of the high-level motivations for pursuing this work is that it has potentially profound impact for education, allowing next-generation educational software to benefit from the motivating power of scenarios and to be generated specifically for the learning profile, goals, and progress of particular learners. It is also potentially the case that experiences generated in this manner, and based on these models, could more effectively reach audiences that have grown up with games as a primary media form.