Survey of How Human Players Divert In-game Actions for Other Purposes: Towards Human-Like Computer Players

Achieving true believability in artificial intelligence requires moving beyond the optimization of primary game objectives to incorporate the nuanced, non-functional actions characteristic of human players. While traditional game agents are designed to maximize efficiency toward a specific goal, human players frequently divert their actions to communicate, express emotion, or engage in social signaling. By classifying these behaviors into seven distinct categories—including warning, notification, and provocation—it becomes clear that these seemingly irrational actions serve as vital social tools, particularly when formal in-game communication channels are restricted or inefficient.

Multi-agent Q-learning experiments demonstrate that these complex behaviors can emerge naturally within artificial systems. When agents are placed in environments where traditional communication is unavailable, they develop action substitution strategies to convey intent, effectively mirroring human social dynamics. These findings indicate that non-objective-oriented behaviors, such as greetings or displays of self-satisfaction, are not merely noise but are essential components of realistic gameplay. Integrating these emergent behaviors into agent design significantly enhances the entertainment value and perceived intelligence of computer-controlled characters.

This research bridges the gap between deep reinforcement learning and the practical requirements of human-computer interaction in virtual environments. By grounding the study in established frameworks for bot believability and Turing test applications, the analysis confirms that the future of game AI lies in the simulation of human-like unpredictability. Ultimately, the development of sophisticated, human-like computer players depends on the successful implementation of these indirect, expressive actions, which transform artificial agents from simple task-solvers into dynamic, relatable participants in the gaming experience.