Research Ethics & IRB

The umbrella IRB classifies every research activity on Uedu into three tiers. This structure was built so that participating teachers and collaborating researchers can quickly determine whether a planned analysis is already covered or requires a separate ethics submission.

Detailed activity catalogs (retrospective vs. newly collected; AI dialogue, forum discussion, wearable sensing, IoT environmental sensing, screen / webcam recording, learner-trait surveys) are listed in the authoritative protocol summary.

Consent is not a single checkbox. The umbrella protocol provides six distinct consent forms, each scoped to a specific data category and signed only when a participant joins the corresponding activity:

1. Platform learning-trace research — newly collected platform data and incidental capture from instructor-side recording (online signature).
2. Wearable physiological data — Garmin / Apple Health / Google Health Connect-derived HRV, sleep, stress, and activity indicators (online, opt-in).
3. Uedu Brain multimodal neural sensing — EEG + fNIRS + PPG in-lab sessions (paper signature, Tier B).
4. Gender-diversity survey — SGBA-Plus questionnaire instruments.
5. Semi-structured interviews — qualitative interview participants (Tier B, paper signature).
6. Expert focus groups — researchers and educators in TEL / LAK adjacent fields.

All consent is granular, revocable, and re-disclosed at the point of feature use. Participants may withdraw individual data streams (e.g., revoke wearable sharing while keeping platform-trace participation) and may request data deletion through the channels described in the privacy policy.

The lab takes the position that, in physiologically- and behaviorally-instrumented learning environments, procedural consent is necessary but not sufficient. The platform is therefore built around four architectural safeguards:

These safeguards are system-enforced rather than reliant on individual researcher discretion — a core design principle of the umbrella protocol.

The platform operates under the Human Subjects Research Act of Taiwan (人體研究法) and the Personal Data Protection Act (個人資料保護法), with NTU-REC providing the institutional ethics review for this protocol.

While the legal jurisdiction is Taiwan, the lab actively positions its research in conversation with international frameworks that increasingly govern emotion inference, biometric processing, and AI in educational contexts:

• EU AI Act — Article 5(1)(f) categorically prohibits AI systems that infer emotions of natural persons in educational and workplace settings, with narrow carve-outs only for medical or safety purposes. Importantly, neither participant consent nor the Act's scientific-research exemption (Article 2(6)) cures an Article 5 prohibition. Recital 18 expressly excludes the inference of physical states such as pain or fatigue — citing pilot and driver fatigue-monitoring as illustrative non-prohibited use. The lab's physiological-sensing components (PALM, Uedu Fit, Uedu Mind) are designed within this physical-state scope: outputs describe autonomic regulation (HRV-derived stress index, sleep quality, activity load), not affective categories such as happiness, anger, or anxiety. See the PhysioNeuromics dimension note for the working principle.
• GDPR principles — granular consent, purpose limitation, data minimization, and the right to withdraw are operationalized at the platform level, irrespective of participant residency.
• OECD AI Principles and UNESCO AI in Education guidance — informing the agency-centered design above.

The lab welcomes inquiries from European, North American, and Asian collaborators wishing to discuss compliance posture for cross-jurisdictional studies.