Abstract:Recent Computer Role Playing Games such as Bethesda’s The Elder Scrolls V: Skyrim and Nintendo’s The Legend of Zelda: Breath of the Wild, have entranced us with their expansive, complex worlds. However, the Non-Player Characters (NPCs) in these games remain stale and lackluster outside of scripted events. This is, in part, because game engines generally do not simulate emotions in their NPCs while they wander in the world. Wouldn't these games be much more interesting, potentially even more re-playable, if NPCs reacted more appropriately to the situations they find themselves in?To be able to do this, designers need an engine that models emotion, based on inputs available in the game world and from other designer-defined character elements such as personality, goals, and mood. A full-fledged cognitive architecture could fulfill this task, but it would likely be much too inefficient for use in a real-time environment like a game.There are many psychological models of emotion but only a few have been explored for video game applications. A game requires an emotion engine which generates believable results to enhance NPC agency and player engagement. Unlike AI agents and simulations of cognitive psychology theories, an emotion engine for games does not need to be correct or even justifiable. This enables the exploration of a variety of emotion theories that have not been actively considered for games. One such theory is Plutchik's psychoevolutionary synthesis. He proposes a method of organizing emotions into a cone, where the intensity of an emotion increases as one moves up the sides. It also postulates that primary emotions in the model can be arranged in opposing pairs and that other emotions can be composed from the primary emotions and their intensities. This allows for greater flexibility in the number and type of emotions to include, whereas most models that have been used before define a closed set of emotion types—a serious constraint on designer's freedom. A second theory, Lazarus's cognitive appraisal, better describes emotion elicitation and behaviour selection, and appears to integrate well with Plutchik's work.An emotion engine based on simplified versions of psychoevolutionary synthesis and cognitive appraisal is an understudied approach towards emotional NPCs. Together with readily identifiable elements of emotion processing, such as attention and action selection, an engine can be designed and customized to meet the needs of game designers with minimal impact on computational resources.We will present an overview of some existing cognitive architectures and emotion engines followed by a description of key elements in psychoevolutionary synthesis and cognitive appraisal. Next we list some requirements for an emotion engine for NPCs and how our selected emotion theories meet them. Finally, we propose a design and a collection of game-oriented test scenarios to illustrate how our design handles various facets of NPC emotional responses.

Emergent social NPC interactions in the Social NPCs Skyrim mod and beyond

Crosston Tavern: Modulating Autonomous Characters Behaviour through Player-NPC Conversation

Research on Actor-Based Virtual Environment System

Action-based Character AI in Video-games with CogBots Architecture: A Preliminary Report

Managing the Personality of NPCs with Your Interactions: A Game Design System Based on Large Language Models

Design Foundations for Emotional Game Characters

A generative model of group conversation

Start Your EM(otion En)gine: Towards Computational Models of Emotion for Improving the Believability of Video Game Non-Player Characters

Emergence of Social Norms in Generative Agent Societies: Principles and Architecture

New And Surprising Ways to Be Mean. Adversarial NPCs with Coupled Empowerment Minimisation

Exploring the Potential of Conversational AI Support for Agent-Based Social Simulation Model Design

Collaborative Quest Completion with LLM-driven Non-Player Characters in Minecraft

CUIfy the XR: An Open-Source Package to Embed LLM-powered Conversational Agents in XR

Social Conjuring: Multi-User Runtime Collaboration with AI in Building Virtual 3D Worlds

AI & NPC IN GAMES

Modelling Structured Societies: a Multi-relational Approach to Context Permeability

Exploring Data-Driven Components of Socially Intelligent AI through Cooperative Game Paradigms

NPCs as People, Too: The Extreme AI Personality Engine

Construction and effect of relationships with agents in a virtual reality environment

LLM-Based Agent Society Investigation: Collaboration and Confrontation in Avalon Gameplay

Computational Social Science: Agent-based social simulation