Escalate or Suppress? How Human–AI Collaboration Modes Shape the Blame Avoidance Behavior of Chinese Street-level bureaucrats

This project contains the data for a study examining how human-AI collaboration modes and algorithm transparency influence the blame avoidance behavior of street-level bureaucrats. The experimental design manipulates two independent factors—human-AI collaboration mode and algorithm transparency—while measuring their effects on blame avoidance behavior, with civil servants’ level of discretion tested as a moderator.This study employs a 2 (human-AI collaboration mode: specialization vs. ensemble) × 2 (algorithmic transparency: high vs. low) × 2 (discretion: high vs. low) between-subjects design, resulting in eight experimental conditions. We observe variations in blame avoidance behavior under different experimental treatments.The data are anonymized and contain no personally identifiable informaingtion. Access and use of the data are subject to the terms specified in this repository.