An Artificial Emotional Agent-Based Architecture for Games Simulation: Brazil

This research proposes a novel architecture for modeling non-player character (NPC) behavior in digital games by integrating psychological models of personality, mood, and emotion. The primary thesis is that incorporating real-world external events into a game’s internal logic—mediated through artificial emotional agents—significantly enhances the believability and realism of virtual characters. By bridging the gap between external data and internal game mechanics, the architecture aims to create more immersive and engaging experiences for players.

The proposed system utilizes four established psychological frameworks: the Big Five model for personality, the Ortony, Clore, and Collins (OCC) model for emotion, the Pleasure-Arousal-Dominance (PAD) model for mood, and fuzzy logic to process vague, natural-language-based inputs into quantifiable data. This combination allows agents to interpret external stimuli and adjust their performance or behavior accordingly. The methodology involves a simulation based on data from the game FIFA 13, where real-world performance metrics serve as inputs to influence NPC behavior.

Validation was conducted through 2,800 total test iterations, categorized into four distinct scenarios. These tests confirmed that positive and negative external events consistently drive corresponding shifts in agent performance, with higher intensity inputs resulting in more pronounced behavioral changes. The findings demonstrate that the architecture successfully maps external environmental factors to internal agent states. While currently focused on sports simulation, the research concludes that this emotional agent-based framework provides a scalable method for increasing the veracity of NPCs across various gaming genres, ultimately fostering a more dynamic interaction between the virtual environment and the player’s reality.