How to Create AI That Makes NPCs Act with Personality: Insights from the Battle AI of 'Monster Hunter Stories 3' [CEDEC 2026]
At CEDEC 2026, Capcom presented how it designed battle AI for ally NPCs in Monster Hunter Stories 3. The system separates decision-making into two layers: machine learning selects broad action categories such as attack, healing, or defense, while traditional methods choose the specific skill within that category. Planners can create training data while playing by selecting the desired action category for each situation, then combine a base model with character-specific adjustments. For AI character design, the case shows personality as a pattern of judgment rather than just dialogue or appearance.

I like that personality is placed in how the NPC chooses action categories, not only in lines or visuals. A character can feel distinct through how they hesitate, heal, or attack in battle.