Believable Non-Player Characters (NPCs) help motivate player engagement with narrative-driven games. An important aspect of believable characters is their contextually-relevant reactions to changing situations, which emotion often drives in humans. Therefore, giving NPCs "emotion" should enhance their believability. For adoption in industry, it is important to create tool development processes to build NPCs "with emotion" that fit current development practices. Psychological validity-the grounding in affective science-is a necessary quality for plausible emotion-driven NPC behaviours. Computational Models of Emotion (CMEs) are one solution because they use at least one affective theory/model in their design. However, CME development tends to be under documented so that its processes seem unsystematic and poorly defined. This makes it difficult to reuse a CME's components, extend or scale them, or compare CMEs. This work draws from software engineering to propose three methods for acknowledging and limiting subjectivity in CME development to improve their reusability, maintainability, and verifiability: a systematic, document analysis-based methodology for choosing a CME's underlying affective theories/models using its high-level design goals and design scope, which critically influence a CME's functional requirements; an approach for transforming natural language descriptions of affective theories into a type-based formal model using an intermediate, second natural language description refining the original descriptions and showing where and what assumptions informed the formalization; and a literary character analysis-based methodology for developing acceptance test cases with known believable characters from professionally-crafted stories that do not rely on specific CME designs. Development of EMgine, a game development CME for generating NPC emotions, shows these methods in practice.

翻译：可信的非玩家角色（NPC）能够激发玩家对叙事驱动型游戏的参与感。可信角色的重要特征在于其能根据情境变化做出符合语境的反应，而人类通常由情感驱动此类反应。因此，赋予NPC“情感”应能增强其可信度。为便于工业界采纳，需创建符合当前开发实践的“具情感”NPC构建工具流程。心理有效性——即植根于情感科学——是产生合理情感驱动型NPC行为的必要条件。情感计算模型（CME）提供了一种解决方案，因其在设计中使用至少一种情感理论/模型。然而，CME的开发过程常缺乏系统记录，导致其流程看似无系统性且定义不清，这使得复用、扩展或缩放CME组件，以及进行CME间比较变得困难。本研究借鉴软件工程方法，提出三种途径以减少CME开发中的主观性并提升其可复用性、可维护性与可验证性：其一，基于系统性文档分析的方法，通过CME的高层设计目标与设计范围（这两者关键影响CME的功能需求）选择其底层情感理论/模型；其二，一种将自然语言描述的情感理论转化为基于类型的正式模型的转换方法，该方法使用对原始描述进行精炼的中间层二次自然语言描述，明确说明形式化过程中的假设内容及位置；其三，基于文学角色分析的方法，利用专业创作故事中已知的可信角色开发验收测试用例，且不依赖特定CME设计。通过EMgine（一款用于生成NPC情感的游戏开发CME）的开发实践，展示了上述方法的实际效用。