The Effect of Power Matching on Trust in Human-Agent Collaboration: Task Difficulty as a Boundary Condition
Description
This dataset contains data from two experimental studies examining how the match between users' personal sense of power and agents' power roles (assistant, partner, leader) influences collaborative outcomes in human-agent visual search tasks. The dataset includes behavioral measures (suggestion adoption rates), subjective ratings (trust, perceived agent power), individual difference measures (sense of power, negative attitudes toward agents), and experimental condition assignments. Ethics approval: Ethics Committee of the Centre for Psychological Science, Shaanxi Normal University. Study 1 employed a 3 (agent role) between-subjects design (N = 194, 86.1% female, Mage = 20.26). Study 2 employed a 3 (agent role) × 2 (task difficulty) mixed design (N = 146, 87.7% female, Mage = 20.45). Both studies used a visual search task with 60 trials. The agent was 100% accurate. Primary DVs: suggestion adoption rate and subjective trust. Variables: ID, Gender (1=male, 2=female), Age, AgentRole (1=assistant, 2=partner, 3=leader), PriorExperience (1=yes, 2=no), SenseOfPower (1-7, α=.72), NARS (1-5, α=.72), PerceivedAgentPower (1-7, α=.88), AdoptionCount (0-60), AdoptionRate (0-1), Trust (1-7, α=.91). Study 2 adds TaskDifficulty (1=high/1500ms, 2=low/2500ms). Procedure: Participants completed scales, 10 practice trials, then 60 trials. Each trial: fixation (500ms) → target (500ms) → search page (2500ms in Study 1; 1500ms high/2500ms low in Study 2) → count report → agent suggestion → accept/reject → role feedback → ITI (500ms). Post-task questionnaires completed. Role Manipulation: Assistant - user final authority. Partner - shared authority. Leader - agent final authority. Manipulation validated by perceived power differences. Data Usage: Use GLM (binomial) for AdoptionRate and linear models for Trust. RSA recommended for congruence effects. Note: behavioral and subjective trust may dissociate; matching effects only under low difficulty; partner role shows broad adaptability. Citation: Authors to be supplemented. Title: Parallel Outperforms Complementary Matching in Human-Agent Collaboration: The Boundary Role of Task Difficulty and the Behavioral-Subjective Dissociation of Trust. Contact: Corresponding author.
Files
Steps to reproduce
Participants first completed the Sense of Power Scale (Anderson et al., 2012) prior to the experimental task. They were then introduced to the visual search task and completed 10 practice trials to familiarize themselves with the interface and procedure. Each participant was randomly assigned to one of three agent role conditions (Assistant, Partner, or Leader), which remained fixed throughout the session. The formal experiment consisted of 60 trials. Each trial proceeded as follows: a fixation cross was presented at the center of the screen for 500 ms, followed by a target image displayed for 500 ms. A search page containing 30 randomly arranged item images then appeared for 2500 ms, during which participants searched for all images belonging to the same category as the target. After the search page disappeared, a response screen was presented, and participants used numeric keys to report the number of targets they had found. The agent then displayed a suggested number, and participants pressed a key to indicate whether they accepted or rejected the suggestion based on their own judgment. Role-specific feedback was subsequently presented according to the participant's assigned agent role. Each trial was followed by a 500 ms inter-trial interval before the next trial began automatically. The entire session lasted approximately 30 to 40 minutes. Upon completion of the 60 trials, participants completed post-task questionnaires on 7-point Likert scales, including measures of trust and perceived agent power. The primary behavioral dependent variable, suggestion adoption rate, was calculated as the number of trials in which participants accepted the agent's suggestion divided by the total number of trials (60).
Institutions
- Shaanxi Normal UniversityShaanxi, Xi'an