Multimodal Dataset of Smartphone Touchscreen Kinetics and Inertial Sensor Logs for Mindful and Mindless Scrolling
Description
This dataset provides high-frequency multimodal behavioral logs captured during smartphone usage, specifically focusing on the distinction between mindful and mindless scrolling behaviors on social media. The data was collected to facilitate research in Human-Computer Interaction (HCI), behavioral biometrics, and digital wellbeing. The dataset comprises logs from 20 participants. To ensure hardware consistency and eliminate sensor variance across different devices, all data was recorded using a standardized research smartphone. Interactions were captured using a custom-built logging application. Participants performed two distinct sessions: a "mindful" session where they were given specific search tasks, and a "mindless" session involving free-form, dissociative scrolling through a TikTok feed. The repository includes: 1. Touchscreen Interaction Logs: Detailed records of interaction events (CLICK and SCROLL) including relative scroll distances (ScrollDeltaX and ScrollDeltaY) and the active application package name. 2. High-Frequency Inertial Sensor Data: Synchronized 3-axis accelerometer and gyroscope logs captured at a combined sampling rate of approximately 105 Hz, providing precise physical orientation and movement data. 3. Participant Metadata: Anonymized demographic information, including age, gender, and handedness (dominant hand). 4. Data Dictionary: A comprehensive guide explaining the schema, units, and column headers for all provided CSV files. This dataset is suitable for researchers interested in: - Developing machine learning models to detect cognitive states or "mindless" behavior in real-time. - Analyzing behavioral biometric patterns through scroll deltas and sensor kinetics. - Studying the physical kinetics of social media interaction on high-frequency data.
Files
Steps to reproduce
1. Software Setup The data collection was performed using a standardized Android smartphone to eliminate hardware-induced variance in sensor sensitivity and screen latency. The device was equipped with a custom-built logging application, designed to capture high-resolution interaction data. 2. Participant and Ethics A total of 20 participants were recruited for this study. Prior to the data collection, each participant was briefed on the research objectives. 3. Experimental Procedure Each data collection session was divided into two distinct scenarios using the TikTok application: - Mindful Scrolling Session: Participants were assigned specific search tasks, inducing a goal-oriented and cognitively active state. - Mindless Scrolling Session: Participants were encouraged to engage in "mindless" or passive scrolling through the TikTok "For You" feed without a specific objective, aiming to capture dissociative usage patterns. 4. Data Acquisition During both sessions, the logger operated in the background, generating two types of CSV log files per session: - Touch Logs: Capturing event types (CLICK and SCROLL), relative scroll deltas (X and Y), and the active application package name. - Sensor Logs: Capturing synchronized 3-axis accelerometer and gyroscope data. 5. Post-Processing and Anonymization The raw log files were extracted and organized into a hierarchical folder structure. All files were fully anonymized by replacing participant identities with unique codes. Any personal identifiers or account-specific information encountered during the log cleaning process were removed to ensure total participant anonymity before publication in this repository.
Institutions
- Binus UniversityJakarta, Jakarta