Evaluating the Impact of LLM-Assisted Natural-Language Commands on Gameplay and Player Experience in The Last Signal
Year: 2026
Type: article
Abstract: Conventional command interfaces require players to remember predefined wording, which can interrupt play when an intended action is expressed differently. This study examines how an LLM changes command-based gameplay in The Last Signal, a Unity puzzle-adventure game in which players direct a robot named PICO. Twelve participants completed the first two rooms with fixed commands and LLM-assisted natural-language commands. The result screen recorded completion time and command use, while questionnaires measured usability, workload, expressive freedom, perceived control, and overall experience. With LLM assistance, mean completion time decreased from 20.75 to 14.58 minutes, total commands from 27.50 to 20.92, and unrecognized commands from 8.08 to 3.42. Participants also reported greater usability and expressive freedom, a lower workload, and an improved overall experience. Perceived control decreased slightly, but this difference was not statistically significant. Interviews indicated that familiar language and PICO's responses strengthened interaction and cooperation, although unexpected interpretations created uncertainty. The findings show that LLM-assisted commands can improve task efficiency and player experience when players receive clear feedback and can easily correct misunderstood instructions. This within-subject comparison evaluates both command performance and the experience of cooperating with PICO. It provides evidence of the value of flexible commands in a constrained puzzle game.
Author: Sitong Wu
Institution: Fuzhou University
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7225744918