WEBVTT

1
00:00:00.700 --> 00:00:13.070
This is Uedu — an AI-mediated learning platform developed by Assistant Professor Chia-Kai Chang, running in production across twenty-nine higher-education institutions in Taiwan.

2
00:00:13.170 --> 00:00:19.684
Over four thousand three hundred learners, and more than four hundred and twenty thousand learning interactions.

3
00:00:19.784 --> 00:00:26.442
Here is what you can run in your own course — with nothing to install, and no code to write.

4
00:00:27.292 --> 00:00:36.950
The idea is simple: you write the teaching behaviour, your students work with it inside the course, and afterwards, you see the kind of thinking that happened.

5
00:00:37.050 --> 00:00:43.852
Most AI teaching tools stop after the first step. This platform exists for the third.

6
00:00:44.702 --> 00:00:47.520
Step one: build the assistant.

7
00:00:47.620 --> 00:00:58.670
Each course gets its own AI assistant. You describe its persona and rules in plain language — how it should answer, what it should refuse, which sources it should prefer.

8
00:00:58.770 --> 00:01:09.676
Upload your own PDFs, slides and documents, and the assistant answers from them — citing the file and the page it used, so students can check every claim against your material.

9
00:01:09.776 --> 00:01:17.130
Optional Socratic and debate modes make it lead with questions, or argue a position your students have to respond to.

10
00:01:17.230 --> 00:01:24.584
And students can build assistants with the same editor. In several courses, that is the assignment itself.

11
00:01:25.434 --> 00:01:28.732
Step two: run the course around it.

12
00:01:28.832 --> 00:01:36.090
Generate quiz and worksheet drafts from your material, then edit them. Nothing reaches students without your review.

13
00:01:36.190 --> 00:01:42.392
Discussion forums can score semantic quality and peer response, rather than post count.

14
00:01:42.492 --> 00:01:50.206
In-platform surveys open and close on schedule — and research consent, where it applies, is versioned and withdrawable.

15
00:01:50.306 --> 00:01:58.524
Class recording captures audio, screen or webcam, with transcripts generated afterwards. Sharing stays off unless you turn it on.

16
00:01:58.624 --> 00:02:08.930
There is also Aida, an independent learning companion that works across a student's whole programme — deliberately Socratic: it asks, rather than answers.

17
00:02:09.780 --> 00:02:12.982
Step three: see how students actually think.

18
00:02:13.082 --> 00:02:19.044
Usage counts tell you who showed up. These tell you what kind of thinking happened.

19
00:02:19.144 --> 00:02:26.402
Dialogue is classified by Bloom's taxonomy and tracked across the semester — per student, and per class.

20
00:02:26.502 --> 00:02:33.664
Learning portfolios gather each student's conversations, assessments and participation into one exportable view.

21
00:02:33.764 --> 00:02:45.870
And for work that needs a measured baseline: twenty-three research-grade cognitive tasks with millisecond timing, plus three open psychometric instruments whose results belong to the student — not your gradebook.

22
00:02:46.720 --> 00:02:49.586
Throughout, you keep control.

23
00:02:49.686 --> 00:03:00.352
A free default model is provided — or your course can run on your own OpenAI, Anthropic or xAI key. Background analysis never runs on your key.

24
00:03:00.452 --> 00:03:08.166
In blind model comparison, students vote between two anonymous models before the identities are revealed.

25
00:03:08.266 --> 00:03:14.276
The interface is available in sixteen languages, and students converse in whatever language they write in.

26
00:03:14.376 --> 00:03:20.746
And your data leaves by one door only: de-identified, logged, and gated on an ethics approval.

27
00:03:21.596 --> 00:03:25.662
The platform is free to use, and it is not a commercial product.

28
00:03:25.762 --> 00:03:32.372
Teach with it: run a course on the platform, with no fee and no obligation to take part in any study.

29
00:03:32.472 --> 00:03:41.170
Or research with it: analyse de-identified data under the platform's umbrella ethics protocol — usually without filing your own submission.

30
00:03:41.270 --> 00:03:46.224
The results, and the failures, are published in peer-reviewed venues.

31
00:03:47.074 --> 00:03:53.540
Uedu. You describe how you want to teach. The assistant follows it — and shows you what happened.

32
00:03:53.640 --> 00:04:00.298
Read more, and get in touch, at uedu.tw/international/teaching.