From Wearable Sensors to AI Insights

Why this tutorial · Why now

Three developments have changed who can build educational sensing infrastructure.

Three developments over the past 24 months have changed who can build educational sensing infrastructure: consumer-grade wearables now expose passive biosignals through documented APIs, AI-assisted development tooling has lifted single-developer throughput on systems work, and AI-assisted testing methodologies have made it tractable to validate AI behavior at the scale at which AI-built systems can now be deployed.

A single principal investigator can now build, validate, and operate the kind of multimodal sensing pipeline that previously required a ten-person engineering team. The remaining bottleneck is no longer engineering capacity — it is measurement methodology, AI behavior testing, and IRB-governed data handling. This tutorial teaches that methodology.

Each technical section pairs a methods-first treatment of the relevant literature with a worked example from Uedu, a deployed multi-tenant AI tutoring platform serving over 3{,}000 students across nine universities under a single umbrella IRB approval (NTU-REC 202507EM058). The platform is a worked example — not the answer. Attendees who prefer to build from scratch take the methodology home.

01

Build it yourself

Edge-to-cloud biosignal ingestion architecture, with UeduPAD (built with AI-assisted development) as a worked example. Decision flowcharts attendees can apply at home.

02

Verify it yourself

A six-layer AI testing rubric covering code, pipeline, behavior, guardrail, governance, and drift. Live walkthrough of an LLM-as-Judge evaluation harness.

03

Govern it responsibly

Redacted SOP templates from a multi-institutional umbrella IRB approval (NTU-REC 202507EM058). Cross-jurisdictional considerations: Taiwan PIPA, Japan APPI, EU GDPR.

§ Learning objectives

What you take home.

By the end of the three hours, you should be able to do these five things on your own, for your own research questions.

01 Objective

Assess whether a consumer wearable can support a given research claim, citing device-specific validation literature.

02 Objective

Design an edge-to-cloud biosignal pipeline with AI-assisted development tooling, with explicit checkpoints for what AI-generated code still requires human verification.

03 Objective

Construct a layered AI-testing rubric (code, pipeline, behavior, guardrail, governance, drift) and instantiate the LLM-as-Judge layer with a versioned prompt.

04 Objective

Integrate physiological, behavioral, and dialogue-trace data under a multimodal schema with explicit time-alignment and consent-state semantics.

05 Objective

Adapt an umbrella IRB SOP template and tiered data-governance policy to your own institutional setting.

§ Schedule · 180 minutes

Three hours, eight segments.

Each technical segment is paired with a methods-first treatment, a worked example from Uedu, and a take-home checklist.

Tutorial Schedule

3 hours

1

Why now: three converging developments

15 min

AI tooling, consumer wearables, and AI testing — and why the bottleneck has shifted to methodology.
2

Consumer wearables validation for research

25 min

Heart rate vs heart-rate variability accuracy against ECG. Device claim envelope by class.
3a

AI-assisted development of edge pipelines

25 min

UeduPAD architecture pattern. Where AI-assisted dev saves time, and where it does not.
3b

AI-assisted testing of AI-built systems

20 min

Six-layer rubric. LLM-as-Judge harness with versioned prompt. Cross-model reproducibility schema.
—

Break

15 min
4

Multimodal fusion and AI analytics

25 min

Physiological × behavioral × dialogue traces under the Educational Omics framework.
5

Responsible AI · IRB · governance at scale

30 min

Cross-institutional umbrella IRB. Redacted SOP. Cross-jurisdictional considerations.
6

Live demonstration

20 min

Real-time BBI streaming + cognitive task + LLM-as-Judge scored live.
7

Open problems and Q&A

15 min

Honest treatment of what is still not solved.

§ CISOSE 2026 federation

How this tutorial maps to eight conferences.

CISOSE 2026 federates eight constituent conferences. This tutorial covers six substantively, with one cross-cutting and one out of scope. Cross-track integration is the structural reason CISOSE is the right venue for this content.

Tutorial coverage by federated conference

●●● core · ●● substantive · ● adjacent

Federated conference	Tutorial segment	Coverage
Service-Oriented Systems Engineering	Overall pipeline framing · §3a · §4
AI Testing & Quality Assurance	§3b · six-layer rubric
Responsible AI	§5 · IRB, governance
Intelligent Mobile Computing	§3a BLE edge · §6 live demo
Big Data & Machine Learning	§4 fusion schemas
Smart Cities & IoT	§3a sensing pattern
Cyber-Intelligence (overall)	Cross-cutting framing in §1
Decentralized Apps / Blockchain	Out of tutorial scope	—

§ Speakers

Two presenters.

Both presenters will be on-site in Fukuoka, leading complementary segments drawn from their respective technical and pedagogical expertise.

Lead presenter

0000-0003-2575-2738

Chia-Kai Chang (張家凱)

Assistant Professor, Center for General Education

National Central University, Taiwan

[email protected]

Founder and principal investigator of the Educational Omics Lab. Builds and studies a multi-tenant AI tutoring platform (Uedu) in production at nine universities. Recent work includes L@S ’26 (cross-disciplinary cognitive engagement at scale), L@S ’26 WiP (PALM — physiologically-aware language models), IEEE EMBC ’26 (C-GRASP), and ICMET 2025 (Educational Omics Data Lake). Holds an IRB umbrella approval for multimodal educational research.

Leads in this tutorial

§1 Why now §2 Wearables validation §3a AI-assisted dev §3b AI-assisted testing §6 Live demo

Co-presenter

0009-0007-3474-8489

Kuei-Hao Li (李奎皓)

Ph.D. Candidate, Interdisciplinary Doctoral Program

National Tsing Hua University, Taiwan

Co-founder of the Uedu platform with a research background in digital learning and AI-assisted instruction in higher education. Co-author on the team's recent publications including L@S ’26, L@S ’26 WiP (PALM), and ICMET 2025 (Educational Omics Data Lake). Focus: pedagogical design, cross-institutional deployment, and engagement-pattern analysis of AI-assisted learning systems.

Leads in this tutorial

§4 Multimodal fusion §5 Responsible AI · IRB · governance

§ Companion publications

Four anchor papers behind this tutorial.

The tutorial draws on, and points to, our team's recent peer-reviewed work. Each is positioned alongside the segment in which it is used as a worked example.

IEEE EMBC 2026

C-GRASP: Clinically-Grounded Reasoning for Affective Signal Processing

C.-L. Cheng, T.-C. Lin, C.-K. Chang

Anchor reference for §6 live demo.

ACM L@S 2026

AI Teaching Assistants at Scale: Cross-Disciplinary Patterns of Adoption and Cognitive Engagement Across Hundreds of University Courses

C.-K. Chang, K.-H. Li

Anchor reference for §4 fusion + §5 governance.

ACM L@S 2026 WiP

PALM: Scaling Physiologically-Aware AI Tutoring Through Consumer Wearables and Large Language Models

C.-K. Chang, K.-H. Li

Anchor reference for §3a edge pipeline + §4 fusion.

ICMET 2025

Designing an Educational Omics Data Lake: A Multimodal Infrastructure for Technology-Enhanced Learning

C.-K. Chang, K.-H. Li

Anchor reference for §4 Educational Omics framework.

§ Tutorial materials

Bring this home in any form you want.

The proposal is available now. Slides, code references, the LLM-as-Judge prompt and rubric, a redacted IRB SOP template, and the AI-testing harness scaffold will be linked from this page closer to the conference date.

Tutorial proposal (8-page IEEE format)

The full proposal includes abstract, schedule, content per segment, CISOSE-coverage mapping, and references.

Pending invited-tutorial confirmation

Slide deck

Forthcoming. Approximate release: 2 weeks before the tutorial date.

Code references & LLM-as-Judge rubric

Forthcoming. UeduPAD reference architecture, the four-dimension LLM-as-Judge rubric, and a minimum-viable AI-testing harness scaffold.

Redacted IRB SOP template

Forthcoming. Recruitment, consent versioning, biosignal handling, data egress, and instructor-side leakage controls — derived from NTU-REC 202507EM058.

§ Venue

Fukuoka, Japan.

IEEE International Conference on Cyber Intelligence and Software-Oriented Service Engineering (CISOSE 2026). July 27 – 30, 2026, Fukuoka, Japan.

https://cisose.fit.ac.jp/2026/

Conference info

Conference

CISOSE 2026

IEEE International Conference on Cyber Intelligence and Software-Oriented Service Engineering

Dates

July 27 – 30, 2026

Location

Fukuoka, Japan

Language

English

Registration

Registration is handled through the CISOSE 2026 main conference website. Tutorial attendance is included with full-conference registration.