Adaptive Horror Games

• How can we identify individual player fears using only gameplay data without explicit questionnaires?
• What machine learning techniques are most effective for modeling player fear responses in real-time?
• How can adaptive systems track horror intensity and dynamically moderate the use of fear-inducing elements?
• What role does virtual reality play in amplifying or modifying fear responses compared to traditional games?
• How can we balance adaptation to player fears while maintaining game difficulty and engagement?

This research develops methods for creating adaptive horror games that can personalize frightening experiences based on individual player characteristics. The approach combines player modeling techniques with machine learning to identify what specific horror elements trigger fear responses in different players.

A core component is the fear model, which maps gameplay observations to horror elements that provoke fear in players. The model uses machine learning techniques to analyze player behavior in response to different horror events, learning to predict which elements — such as darkness, apparitions, unknown sounds, or confined spaces — will be most effective for each individual player.

The research also explores adaptive agent-based systems that can track the horror intensity experienced by players throughout the game and dynamically moderate the use of fear-inducing elements. This enables the system to maintain appropriate pacing and prevent desensitization, while continuously adapting to changing player responses during gameplay.

Virtual reality technology plays a significant role in this research, as VR environments can amplify immersion and psychological presence, potentially intensifying fear responses. The project investigates how VR-specific elements and interaction methods can be used to create more effective adaptive horror experiences while maintaining real-time performance and player engagement.

Watch two gameplay sessions of the adaptive VR horror game prototype developed for this research project, alongside visualizations showing how player fear responses changed during gameplay:

Session Example 1: Stable Fear Response

Gameplay Recording

This example shows a participant with stable fear responses throughout the game session, with no considerable fear changes detected by the model during gameplay.

Session Example 2: Evolving Fear Response

Gameplay Recording

This example demonstrates a participant showing gradual transitions in fear responses during the game session, with the model detecting and adapting to changing fear patterns over time.