Data for: Virtual Agent Appearance, Social Perception, and Partner Preference in Immersive Environments
Description
This dataset contains the raw episode-level data from four controlled experiments examining how virtual agent appearance (human-like vs. robot-like) and behavior (prosocial vs. individualistic social value orientation, or high vs. medium task efficiency) shape social perception (warmth and competence) and cooperation preferences in an immersive virtual reality environment. Participants were recruited via Prolific and completed a VR-based cooperative task, the Coins game, using a VR headset. After each episode, participants rated their AI co-player on warmth and/or competence, and after every two episodes, they reported a stated cooperation preference between the two co-players. The dataset includes four CSV files, one per study reported in the associated manuscript. Study1_Data.csv contains data from Study 1 (N = 52 participants), which manipulated agent appearance and prosocial/individualistic behavior to examine effects on warmth perceptions and cooperation preference. Study2_Data.csv contains data from Study 2 (N = 52 participants), which manipulated agent appearance and task efficiency (high vs. medium) to examine effects on competence perceptions and cooperation preference. Study3_Data.csv contains data from Study 3 (N = 53 participants), a replication of Study 1 in which the human-like avatar's gender was changed, to test whether the appearance-warmth effect observed in Study 1 was attributable to avatar gender rather than the human-versus-robot appearance contrast. Study4_Data.csv contains data from Study 4 (N = 62 participants), a replication of Study 1 in which the robot-like avatar's design was changed from a friendly, approachable robot to a mechanically harsh, non-cute robot, to test whether the warmth advantage observed for robot-like agents in Studies 1 and 3 depended on the specific aesthetic properties of the robot design rather than the human-versus-robot distinction itself. This file also includes participant responses to nine post-study perception items for each of the four agent variants (Adam Prosocial, Adam Individualistic, Jammo Prosocial, Jammo Individualistic). Each file contains episode-level records, with one row per episode per participant. Variables include participant ID, pair and episode sequence numbers, agent type, appearance, and behavioral condition, player and AI scores, warmth and/or competence ratings, stated cooperation preference, bonus payment, participant demographics (age, gender, VR experience), and coin-collection counts by color and player.