PE-HRI-temporal: A Multimodal Temporal Dataset in a robot mediated Collaborative Education

PE-HRI-temporal is a multimodal temporal dataset from the JUSThink robot-mediated collaborative learning activity. It contains team-level behavioral time series and learning outcome data from 34 teams of two children aged 9-12. The dataset was created by Jauwairia Nasir at École Polytechnique Fédérale de Lausanne.

Use Cases

• Modeling student engagement based on log actions, speech behavior, affective states, and gaze patterns.
• Analyzing the evolution of team collaboration over time using 10-second windowed behavioral features.
• Predicting team-level learning outcomes like absolute and relative learning gains from temporal interaction data.
• Studying the relationship between affective states and performance in a robot-mediated educational setting.

Strengths

• Data is collected from 68 participants organized into 34 teams, providing a basis for team-level analysis.
• Includes 5,048 time windows of 10 seconds each, enabling fine-grained temporal analysis of behavior.
• Captures 52 features per time window, combining non-incremental and incremental versions of 26 behavioral metrics.
• Links temporal behavioral data to concrete learning outcome metrics like T_LG_absolute and T_LG_relative.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the primary behavioral file is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Jauwairia Nasir, École Polytechnique Fédérale de Lausanne

Collection Method

Collected during the JUSThink robot-mediated collaborative learning activity.