How Players Speak to an Intelligent Game Character Using Natural Language Messages

Natural language interaction serves as a transformative tool for enhancing player engagement and the perceived agency of non-player characters within digital environments. By analyzing 186 message logs from a Wizard of Oz study involving 18 students in Minecraft, it becomes clear that players project significant anthropomorphic expectations onto AI companions. These interactions are heavily influenced by the user’s prior experiences with robotics, social norms, and existing game mechanics, which collectively shape how individuals conceptualize and communicate with intelligent agents.

Players predominantly utilize short, imperative phrases to issue commands, often avoiding complex linguistic structures due to uncertainty regarding the system’s comprehension capabilities. A primary challenge in these interactions is the frequent lack of shared context, which leads to ambiguity concerning object specificity, location, and task boundaries. To bridge this gap, players often rely on in-game demonstrations and physical actions to clarify their intent, highlighting a reliance on the virtual environment to supplement textual inputs.

Successful implementation of natural language interfaces requires designers to prioritize clarity and feedback. Because players struggle to form accurate mental models of an AI’s internal state, systems must provide transparent communication regarding learning progress and decision-making processes. Features such as contextual autocompletion, customizable command menus, and prescriptive feedback mechanisms are essential to reduce user hesitation and improve usability. Ultimately, while natural language input significantly increases the perceived aliveness of game characters, its effectiveness is contingent upon the agent’s ability to interpret environmental context and provide clear, actionable responses that align with player expectations.