Spotting the Difference: Identifying Player Opponent Preferences in FPS Games

This study investigates player perceptions of artificially intelligent (AI) opponents in first-person shooter (FPS) games, specifically focusing on the ability of players to distinguish between human opponents and computer-controlled bots. The research aims to identify which behavioral cues are most effective for this classification and to understand player preferences regarding opponent types. By analyzing these interactions, the study seeks to inform the development of more engaging and realistic game AI.

The methodology involved a user study with 20 participants who engaged in deathmatch-style sessions using Quake III: Arena. Participants played both one-on-one and four-player matches against a mix of human and bot opponents. Following the gameplay, participants completed questionnaires to rate the "humanness" of their opponents on a seven-point Likert scale and to evaluate the utility of 11 specific behaviors—such as aiming, camping, and fleeing—in identifying opponent types.

Key findings indicate that while players can generally distinguish between humans and bots, the accuracy of these identifications is lower than expected, particularly in four-player matches where uncertainty increases. Participants demonstrated a clear preference for human opponents over bots, even when they misidentified a human as a bot. The study concludes that reactive behaviors—specifically how an opponent responds to player actions, such as detecting, pursuing, or fleeing—are the most significant indicators of humanity. These results suggest that future AI design should prioritize dynamic, responsive behaviors over static skill-based adjustments to better mimic human play styles and enhance the overall player experience.