A Hybrid Dataset for Studying Human Trust Dynamics in Sequential Human-Robot Collaboration

This dataset accompanies the paper "A Hybrid Dataset for Studying Human Trust Dynamics in Sequential Human-Robot Collaboration" (under review at Scientific Data). It provides a comprehensive resource for analyzing trust evolution, human and robot decision-making, and sequential outcomes across three complementary experimental modalities: LLM-simulated, VR-based human-in-the-loop, and real-world human-robot collaboration. Dataset Content:The dataset contains temporally aligned records of sequential human-robot collaborative tasks. For each participant (or simulated agent), data are collected for ten consecutive tasks, with each record including: trust: Reported trust ratings before each task state: Ground-truth environment state (“threat” or “no threat”) robot_observation: Robot’s observation per task robot_action: Robot’s recommended action/advice (“call for support” or “enter directly”) human_action: Human (or simulated) action per task task_result: Task reward (+5 or -5) Experimental Modalities: LLM-Simulated data: Large language model (GPT-4) simulates the human collaborator, enabling large-scale generation of trust and decision trajectories under the unified protocol. VR-Based data: Real human participants interact with a simulated robot in an immersive virtual reality environment, following the same sequential protocol. Real-World data : Human participants collaborate with a quadruped robot in a physical indoor environment that mirrors the simulation setting. File Organization:Each modality is provided as a separate JSON file: LLM_Simulated_data.json VR_Based_data.json Real_World_data.json Potential Uses: Modeling and prediction of human trust dynamics Analysis of human-robot collaborative behavior Benchmarking algorithms for sequential trust-aware decision making Transfer learning and sim-to-real adaptation Additional Information:Detailed variable definitions, protocols, and sample code for loading and analyzing the dataset are provided in the README file and Supplementary Methods of the associated manuscript. All data are anonymized and provided under the Creative Commons Attribution 4.0 International License (CC BY 4.0). Contact:For questions or requests, please contact the corresponding author at [e-mail].