[Preprint] Start Your EMgine—A Methodology for Choosing Emotion Theories for Computational Models of Emotion

Believable Non-Player Characters (NPCs) are a crucial component of narrativedriven games, which offer rich player experiences (PX). The quality of these NPCs’ interactions with players motivates continuous engagement with the game, and often re-engagement after players stop playing. An important aspect of believable characters is their contextually-relevant reactions to changing situations, including those caused by player actions. In humans, this is frequently driven by emotion. The plausibility of NPCs’ reactions partially depends on their psychological validity. A Computational Model of Emotion (CME), grounded in emotion theories and/or from psychology, is one solution because they can compute which emotion an NPC might experience given the current game state. A CME’s chosen underlying theories should be a consequence of its high-level requirements. We present a methodology that uses a set of emotion theories and a game-oriented CME’s high-level requirements and design scope to produce a grounded judgment of which theories best support the requirements.