Abstract
New and Novice players in competitive MOBAs like League of Legends often experience frustration due to the imbalance of flow, leading to low retention from players, who are unable to learn the complex mechanics. While existing Non-AI and AI tools provide item recommendations based on popularity, they lack the explanations, which can create player dependency. No studies go into detail of whether an AI assistant with explanations can specifically reduce frustration in new and novice players.
This research implements an AI ’Itemization’ Assistant combining a Deep Q-Network model for item recommendations with a large language model for real-time explanations. Using a within-subjects design, 54 participants each played two 1v1 League of Legends matches, one with the AI assistant active and one without. Player experience was measured using the Game Experience Questionnaire and system usability assessed with the System Usability Scale questionnaire.
The assistant significantly reduced frustration and improved overall player experience. All four GEQ sub-scales showed significant improvements in the AI condition. The results suggest that frustration reduction is attributed to the explained by the AI Assistant, independent of practice effects.
Overall system usability was the strongest predictor of frustration reduction, and participants showed a clear preference to explanations, rather than recommendations. Knowledge gain was significant overall, but improvement was concentrated among beginner and intermediate participants, while the majority of those who had never played before, showed no change.