Computational Solutions To The Exponential Problem in Longform Branching Interactive Digital Narratives

该提议是探索计算方法，以解决分支数字叙事中的内容创建的指数问题。分支是一种流行的方式，可以在互动体验中表达玩家的自主性(无论是主要由文本组成的动态小说，还是视频游戏)。在分支叙事中，每次向现有分支添加一个分支，就会以指数级增加作者的工作量。如果一个故事有两个分支，并且每个分支都有两个分支，依此类推，那么在十个选择之后，将需要写入1024个节点。这个指数问题强烈地限制了叙事中可以出现的复合分支的数量。对于这个问题，有许多现有的结构性解决方案可以建模。其中包括在分支后频繁合并的结构，或使用状态跟踪进行延迟分支的结构。这些方法可以在大多数现有的分支叙述中看到。除了这些结构解决方案，还有一些很有前途的计算方法。使用人工智能叙事导演是一种可能性。另一个是为角色和环境设置程序规则，这样叙事就可以出现在一个更沙盒的环境中。其目的是研究分支问题的各种现有解决方案，并探索生成技术如何扩展它们。

The proposal is to explore computational approaches to solving the exponential problem for content creation in branching digital narratives. Branching is one popular way of allowing player autonomy to be expressed in an interactive experience (whether it be a kinetic novel formed mainly of text, or a video game). In a branching narrative, each time a branch is added to an existing branch, it increases the workload of authorship exponentially. If a story has two branches, and each of those branches has two branches and so on, after just ten choices, 1024 nodes would need to be written. This exponential problem strongly limits the amount of compounded branching that can occur in a narrative.There are a number of existing structural solutions to this problem which can be modelled. These include structures that frequently merge after branching, or use state tracking for delayed branching. These approaches can be seen in most existing branching narratives. Beyond these structural solutions, there are also promising computational approaches. Using an AI narrative director is one possibility. Another is setting up procedural rules for characters and environments so narratives can emerge in a more sandbox environment. The intention is to research the various existing solutions to the branching problem, and explore how generative techniques might extend them.