Evaluating the effectiveness of an adaptive game based on player's emotions
Loading...
Date
2025-06-30
Authors
Σαμαντά, Μαρία
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis investigates how facial emotion recognition can be used to create more personalized and responsive gaming experiences through emotion-adaptive design. The central scientific problem addressed is the integration of emotional feedback into real-time gameplay in a way that meaningfully affects the player’s experience. To explore this, a 2D platformer game was developed in Unity that dynamically adapts its mechanics based on the player’s facial expressions. A standalone emotion recognition system was used to classify four emotional states (Happy, Sad, Angry, and Surprised) using webcam input. These predictions were transmitted to the game in real time and used as implicit input to trigger changes in gameplay elements such as enemy speed, obstacle behavior, supportive power-ups, and background music tone. A controlled user study involving 37 participants was conducted to evaluate the system. Each participant played one static and one emotion-adaptive level in randomized order. Data collection included questionnaires, gameplay logs, and EDA signals, which were used to both assess player experience and validate the FER system.
Chapter 2 presents the related literature on affective computing, emotion-adaptive interfaces, emotion recognition techniques, and the role of gamification and platformer games in HCI research. Chapter 3 outlines the design and implementation of the game, including game mechanics, level design, and adaptive features. Chapter 4 describes the experimental protocol and data collection process, covering gameplay sessions, questionnaire design, integration with the FER module, and physiological signal acquisition. Chapter 5 details the preprocessing steps applied to the gameplay logs, EDA recordings, and questionnaire responses, ensuring clean and analyzable datasets. Chapter 6 presents the results of the study, analyzing player experience, awareness of adaptation, performance differences, gender and gamer-type effects, and physiological validation of the FER system.
The results showed that more than half of the participants were able to correctly identify the adaptive level, suggesting that the emotion-driven mechanics, though subtle, were detectible. Subjective ratings indicated small but consistent differences in perceived difficulty and control. Objective data revealed that performance improved in the second level played, regardless of type, due to learning effects. Gender and gaming experience had strong effects on both performance and perception, with female players encountering more difficulty and gamers benefiting more from the adaptive gameplay. EDA signals confirmed increased arousal during the dynamic level and during “Sad” emotion segments, supporting partial validity of the FER system. However, issues such as misclassifications, undetected emotions, and participant fatigue limited overall reliability.
This work contributes to the growing field of emotion-aware game design by demonstrating a functional emotion-adaptive system and offering a framework for its evaluation. Limitations include a relatively small sample, subtle adaptations that were sometimes misunderstood, and technical constraints in emotion detection. Future research could improve recognition robustness by using newer training data and include additional emotion modalities to detect player’s emotion such as posture and EDA. The findings support the potential of emotion-adaptive systems and highlight opportunities for creating more inclusive and emotionally engaging interactive experiences.
Description
Keywords
Emotions, Games, Adaptive games