Reflections of Cognitive Load in Player Discourse: A Linguistic Profiling of Steam Reviews for RPGs

Background: Cognitive load (CL) in digital games is usually measured with lab-based sensors or questionnaires, which are intrusive, costly, and hard to scale. Existing studies rarely examine how players’ own discourse reflects intrinsic, extraneous, and germane load. Methods: We propose the Game-Adaptive Linguistic Profile (G-LP) model, which maps linguistic indicators (e.g., sentence length, entropy, negativity, thematic term density) to the three components of Cognitive Load Theory. Using NLP on 15,000 English Steam reviews from five RPGs (Baldur’s Gate 3, Disco Elysium, Divinity: Original Sin 2, Pathfinder: Wrath of the Righteous, Pillars of Eternity II), we computed game-level CL profiles and compared them via Kruskal–Wallis and Dunn–Bonferroni tests. Results: All three CL profiles differed significantly across games (H_intrinsic = 344.92, H_extraneous = 100.44, H_germane = 186.75, all ps