Virtual Grid Aggression Tasks Reveal Intention-Dependent Cooperation and Affective Responses to Digital Agents

Aggressive and cooperative responses can both be adaptive depending on the intentions and the appearance of the interaction partners. However, few approaches combine dynamic decision-making between aggression and cooperation with controlled and realistic social cues. We developed two virtual grid aggression tasks (GATs) to study how competitive versus cooperative intentions of non-player characters (NPCs) of different appearances shape behavior, affect, and physiological responses.In a screen-based study (Experiment 1; N = 30), participants interacted with geometric NPCs in a grid-foraging task permitting both mutual gain and competitive blocking. In an immersive VR version (Experiment 2; N = 57), NPCs were rendered as life-sized avatars varying in affiliative appearance, via facial manipulation of apparent trustworthiness in stylized male characters, or participant-specific photorealistic “doppelgangers” (NPC avatar styles “T-NPC” and “D-NPC”). Across both studies, participants adjusted cooperation to NPC intention: cooperation increased toward cooperative NPCs and declined toward competitive NPCs. Competitive NPC intention also elicited higher anger evaluations in both environments.In VR, affiliative appearance predicted self-reported identification with NPCs in the hypothesized direction, independent of avatar style. In addition, intimidation responses and stress-related physiological measures differentiated avatar styles: stylized trust-manipulated avatars induced higher intimidation, heart rate, and respiration, and lower heart rate variability, whereas photorealistic doppelgangers elicited lower arousal and greater visual engagement. Trait aggression predicted anger ratings for competitive but not cooperative NPCs. Exploratory moderation effects with gender appeared appeared in specific VR conditions for trait aggression and self-reported identification with NPCs variables.These results support virtual GATs as experimentally controllable platforms for studying aggression and cooperation-related affect in immersive social environments. Thus GATs can be used to investigate interpersonal predictors and the role of identification in dyadic interactions that allow for cooperation and aggressive competition. The presented approach links social decision-making to perceptual, physiological, and identification-based responses within a unified framework.