Technical Summary: MotionMonitor

Problem Statement

While wearable sensing technologies offer potential for quantifying ADHD-related behaviors (such as movement, physiological changes, and attention lapses), existing systems often prioritize data capture or automated detection over the interpretation of that data within everyday educational practice. Current approaches frequently separate sensor traces from classroom context, risking increased burden, stigma, or surveillance rather than providing meaningful support. Furthermore, ADHD identification is inherently multi-informant and context-dependent; the same behavior may be expected in one setting (e.g., playground) and concerning in another (e.g., structured instruction). There is a lack of accessible wearable interfaces that help educators and clinicians translate raw sensor data into interpretable, context-sensitive information that fits real classroom workflows without replacing human judgment or attempting automated diagnosis.

Methodology

The authors present MotionMonitor, a technology probe designed to explore how wearable data can support contextualized interpretation of ADHD-related behaviors. The study employed a formative qualitative design research approach rather than a clinical validation or large-scale deployment.

• System Design: MotionMonitor pairs a student-worn smartwatch (Apple Watch) with an educator-operated mobile application (iOS). The system does not collect continuous, decontextualized data. Instead, it organizes sensing around labeled classroom activities.

  • Workflow: An educator selects a participant, chooses a classroom activity (e.g., independent work, transitions), and starts a timed recording. The watch captures heart rate, step count, wrist temperature, blood oxygen, respiratory rate, and accelerometer data (sampled at 100 Hz).
  • Output: Data is stored locally and exported as CSV files, preserving the relationship between sensor samples, timestamps, participant IDs, and specific activity labels.
  • Constraints: The system explicitly avoids automated diagnosis, ranking, alerts, or behavioral labeling.

• Evaluation: A formative study was conducted with 11 experts (clinicians and educators) with direct experience supporting children with ADHD.

  • Data Sources: Semi-structured interviews (approx. 1 hour each) and classroom observation notes from a specialized ADHD-focused school (16 hours of observation across 4 sessions).
  • Procedure: Participants did not deploy the system in a live classroom setting. Instead, they engaged with the prototype through scenario-based prompts and feature critiques to discuss how wearable data might be collected, labeled, interpreted, and governed.
  • Analysis: Interview transcripts and observation notes were analyzed using thematic analysis (open coding, constant comparison, and memoing) to identify design requirements and interpretive needs.

Key Contributions

1. Design and Implementation of MotionMonitor: A classroom-based accessible wearable monitoring technology probe that organizes sensing around labeled classroom activities, linking physiological and movement signals to specific student tasks.
2. Characterization of Expert Interpretation: Findings detailing how educators and clinicians recognize and judge ADHD-related behaviors. Experts view these behaviors as embodied, multidimensional, and context-dependent, emerging from the interaction between a child's regulatory profile and specific activity demands.
3. Design Implications: A set of five implications for accessible, ethical, and low-burden wearable systems that support sensemaking, multi-informant reflection, student comfort, and responsible non-diagnostic use.

Results and Findings

The study yielded four primary themes regarding the interpretation of ADHD behaviors and the role of wearable technology:

1. ADHD-Related Behaviors are Embodied and Multidimensional: Experts do not view behaviors as isolated symptoms (e.g., just "hyperactivity"). Instead, they interpret movement, speech, attention, emotional reactivity, and sensory-seeking as overlapping forms of regulation. The quality and organization of movement (e.g., intentional vs. disorganized) matter more than raw counts.
2. Concern Emerges Through Context, Bias, and Masking: A behavior is only "concerning" when it is developmentally inappropriate, persistent across settings, or disruptive to participation. Experts noted that visibility is shaped by bias (e.g., gendered expectations where girls may mask symptoms) and that classroom data alone is incomplete without cross-setting perspectives.
3. Monitoring is Collaborative but Fragmented: Current monitoring relies on fragmented tools (rating scales, logs, emails) across teachers, parents, and clinicians. Experts expressed a need for technology that acts as infrastructure for organizing observations across stakeholders, rather than a standalone sensor.
4. Wearable Sensing is Only Useful When Interpretable and Non-Diagnostic: Experts emphasized that raw data streams are insufficient. Data must be:
   • Activity-linked: Tied to specific classroom routines.
   • Low-burden: Easy for teachers to label and review without adding significant workload.
   • Comfortable: Devices must not distract or cause sensory discomfort for the child.
   • Non-diagnostic: Data should serve as prompts for human interpretation, not as automated behavioral verdicts.

Significance and Claims

The paper claims that MotionMonitor serves as a bounded design artifact to explore the conditions under which wearable data can be meaningfully interpreted in neurodiverse educational contexts. Its significance lies in reframing classroom wearable sensing from a "behavior-detection pipeline" to an "activity-linked interpretation infrastructure."

The authors argue that for wearable systems to be accessible and ethical in schools, they must:

• Preserve classroom activity as an integral part of the data record, not optional metadata.
• Summarize embodied patterns rather than isolated sensor events.
• Create shareable evidence records that support multi-informant interpretation.
• Treat wearable comfort and distraction as conditions of data quality, not just usability.
• Use sensor data as prompts for reflection rather than behavioral verdicts to avoid reinforcing inequities or bias.

The study concludes that while sensing capability is necessary, it is insufficient without interfaces that support human sensemaking, contextualize data within daily routines, and respect the social and interpretive nature of ADHD assessment.