AI Literacy Course

AI Literacy Before DeepFake Detection:
A Course on Use, Limits, and Verification

This academic version frames DeepFake detection as one part of a broader AI literacy curriculum. The course first asks what generative AI can help people do, then examines where it fails, and only after that moves into verification, visual forensics, DeepFake detection, and real-world misinformation cases.

Start with AI Capabilities Responsible Use DeepFake & Visual Forensics

⚡ Course guide

The recommended sequence is intentional: enable constructive AI use first, introduce limitations and failure modes second, and then teach verification and DeepFake forensics. This reduces defensive resistance and makes risk education feel like responsible use rather than blanket rejection.

Course framing

This version is designed for university courses, invited talks, cross-disciplinary workshops, and public-interest collaborations. It keeps the evidence base, policy references, and Taiwan fact-checking context visible.

Academic route

Chapter 01: AI capabilities and limits
Chapter 05: responsible use in school, work, and sharing
Chapter 04: visual forensics and DeepFake detection
Chapters 06–07: tools, workflow, and real-world cases

Evidence Base

This curriculum is built on MIT, UNESCO, NIST, Cambridge, WITNESS, and Taiwan fact-checking practice rather than intuition alone.

MIT News / Science

MIT: False news travels faster than true stories

Official MIT summary of the Science paper behind the 6× / 70% / human-not-bot claim.

UNESCO

UNESCO guidance on generative AI in education

Explains why AI literacy must cover privacy, age, validation, and human oversight.

NIST

NIST Generative AI Risk Management Profile

Core course stance on trustworthiness, transparency, verification, and oversight aligns with NIST.

Real incidents for verification practice

These cases appear after the AI literacy framing, so they function as practice material rather than fear-based opening examples. Each image links back to the original report or article for that event.

Tel Aviv meteor-like airstrike case image

Case *1

"Meteor-like airstrike on Tel Aviv" AI video

Crisis footage spreads because it looks urgent and cinematic. That makes it a strong teaching case for synthetic video plausibility.

Source: Taiwan FactCheck Center, 2026-03-26

Case *2

Fake edited "20 years ago Lai speech" clip

Not every false video is fully generated. Some are edited, reframed, and circulated as if they were authentic historical material.

Source: Taiwan FactCheck Center, 2025-03

Case *3

AI-generated "elderly duet" video

Emotionally warm content is often shared with less skepticism, which makes it useful for teaching public verification habits.

Source: Taiwan FactCheck Center, 2025-02

Case *4

Leaked fake political audio

Verification is not only visual. Reposted audio clips and platform-transformed voice content create their own misinformation risks.

Source: RFA / Asia Fact Check Lab, 2024-06-11

Case *5

AI video of insects making flowers bloom instantly

Highly aesthetic “nature magic” clips are useful for teaching how overly perfect motion and timing can still be a red flag.

Source: Taiwan FactCheck Center, 2025-05

Case *6

When an AI tool explains the news incorrectly

This case broadens the lesson beyond video: generative systems can also present wrong or incomplete information as if it were authoritative.

Source: Taiwan FactCheck Center, 2025-03

Visual & Deepfake Forensics