Remote Biology Laboratory Interaction Dataset

The Remote Biology Laboratory Interaction Dataset consists of session-level records collected from AI-enabled remote biology laboratory platforms used in higher education settings. The dataset captures detailed learner interaction, execution behavior, and assessment-related information generated during mobile-based laboratory sessions conducted in real instructional environments. Each record corresponds to a single completed laboratory session and aggregates interaction logs, procedural execution indicators, learning analytics, and contextual attributes observed throughout the session lifecycle. The dataset contains a total of 254,020 records, reflecting diverse learner behaviors across multiple laboratory activities, academic stages, and usage contexts. To ensure realism and applicability, the data exhibits natural imbalance across behavioral features and learning outcome categories, consistent with typical mobile learning and remote laboratory deployments. The target label, Neuro-Integrated Learning Outcome State (NILOS), categorizes session-level learning attainment into four levels: Emergent, Adaptive, Proficient, and Mastery, enabling fine-grained outcome analysis beyond conventional score-based evaluation. Collected attributes span cognitive-behavioral indicators, mobile interaction patterns, laboratory execution metrics, assessment outcomes, temporal progression signals, and contextual learning factors. These heterogeneous features reflect the multifaceted nature of learner engagement in remote experimental environments and support advanced learning outcome modeling and analysis. Prior to analysis, standard quality control procedures were applied, including normalization, missing-value handling, and redundancy reduction, to ensure consistency and reliability across sessions. The dataset is made publicly available through Zenodo to support reproducibility, benchmarking, and further research in mobile learning analytics, remote laboratory assessment, and AI-driven educational systems. It is suitable for multi-class classification, robustness evaluation, interpretability analysis, and federated or distributed learning studies involving large-scale educational interaction data.