AN EMPIRICAL-ANALYTICAL STUDY OF PLAYER PERCEPTION, ENGAGEMENT, AND SATISFACTION IN GAMING USING ARTIFICIAL INTELLIGENCE

Background: Artificial intelligence (AI) techniques are now deeply integrated into modern video game systems. They control non-player character (NPC) behaviour, generate game content, adjust difficulty, detect cheating, and personalise player experiences. However, it is still not well understood which AI features most strongly affect how players engage with and feel about games.
Objectives: This paper investigates how six AI feature dimensions—Adaptive Difficulty (AD), NPC Realism (NPC), Procedural Content Generation (PCG), Anti-Cheat Fairness (ACF), AI-Driven Narrative (AIN), and Personalized Recommendation (PR)—affect Player Engagement (PE) and Player Satisfaction (PS) among video game users.
Methods: A structured questionnaire was administered to 412 active game players. Perceptions of AI features were recorded on a five-point Likert scale. The data were analysed using descriptive statistics, Pearson correlation, multiple linear regression (OLS), one-way ANOVA, and five machine learning (ML) classifiers evaluated using 10-fold stratified cross-validation.
Results: Anti-Cheat Fairness received the highest mean perception score (M = 4.28, SD = 0.38), followed by Adaptive Difficulty (M = 4.12, SD = 0.41). All AI features showed significant positive correlations with both Player Engagement and Player Satisfaction (r = 0.29–0.74, p < .01). Regression analysis showed that the combined AI features explained 61.4% of the variance in Player Engagement (R² = 0.614, F(6, 405) = 109.22, p < .001) and 57.8% of the variance in Player Satisfaction (R² = 0.578, F(6, 405) = 93.47, p < .001). One-way ANOVA revealed significant differences in AI satisfaction across experience levels (F(3, 408) = 18.74, p < .001, η² = 0.12). Among classifiers, Gradient Boosting achieved the best accuracy (78.3%) and AUC-ROC (0.91).
Conclusion: AI feature perceptions significantly predict player engagement and satisfaction. Adaptive Difficulty and Personalized Recommendation are the strongest computational drivers of player outcomes. These findings offer concrete design guidelines for AI systems developers and game engineers.