Using Biometrics to Understand AI-Assisted Coding Performance and its Perception - Dataset

Dataset for the paper "Using Biometrics to Understand AI-Assisted Coding Performance and its Perception" by Burelli, P., Calefato, F., Grassi, D., Hristova, M. Y., Novielli, N., Romano, A. A., & Tell, P. This replication package contains data from a controlled experiment with 60 participants (bachelor, master, and PhD students) who performed Java programming tasks with and without AI assistance, conducted at the University of Bari (UniBa) and the IT University of Copenhagen (ITU).The dataset includes: Biometric signals: EEG, GAZE, EDA-HRV (UniBa only) Behavioral logs: keyboard activity, mouse events, GPT interaction logs, oTree experiment events Subjective measures: NASA-TLX workload assessments, demographic and post-experiment questionnaires Code submissions and evaluations: participant Java source files with manual code quality scores Screen recordings of participant sessions Analysis-ready final datasets with extracted features, and normalized NASA-TLX scores Raw data follows the BIDS standard. The repository also includes the full Dockerized experiment infrastructure (oTree, code-server, LSL), analysis scripts (Jupyter notebooks), and task materials. See the README for full details on structure, extraction, and reproduction of the analysis pipeline.