Drucker-Simon AI Literacy Behaviors Dataset

Dataset Description: A Classroom Intervention Study on Improving Generative AI Literacy Among Non-STEM Undergraduates​This dataset originates from a classroom intervention study aimed at enhancing generative AI literacy among non-STEM undergraduate students. Data collection took place from May to September 2025 in a Public Administration undergraduate course at a Chinese university, spanning a four-month period that included three phases: baseline assessment, immediate intervention, and a 3-month follow-up.The core content of the dataset comprises complete process records for 48 participants, including:​Behavioral coding data​ (Sampling frequency: 12 key teaching behavior points recorded per 50-minute class session).​Rubric scoring records​ (Assessing the quality of AI concept explanations using a 15-point scale).​Academic performance correlation data​ (Rate of change in assignment scores before and after the intervention).The data table structure adopts a standard two-dimensional format, with each row corresponding to an observation sample (48 rows in total). Column labels include:Student Anonymous ID (StuNo)Intervention Group (Group: Intervention/Control)Behavior Total Score (Behavior_Total, range 0-24 points)Explanation Quality Score (Explanation_Score, range 0-15 points)Grade Improvement Rate (Grade_Improvement, percentage measure)Regarding data completeness, the collection rate for all core indicators reached 100%. Only three instances of loss to follow-up occurred in the follow-up phase (missing rate: 6.25%), which are marked as NA in the table. Measurement errors primarily stemmed from the double-blind scoring process, with an intraclass correlation coefficient (ICC) ranging from 0.81 to 0.89, which is within an acceptable range.The dataset files include:Teacher_ABC_RawScores.csv (Raw scoring records)Behavior_Coding_Matrix.xlsx (Behavior coding matrix)ICC_Analysis.py (Reliability analysis code)Files are in universal CSV, XLSX, and PY formats, accessible without specialized software. To run the analysis code, a Python 3.8+ environment with libraries such as Pandas and Pingouin is recommended