TwiddleLab / Artificial Intelligence / AI and Us: Using AI Well
Deepfakes and Media Literacy
What you will learn
- A deepfake is AI-made media that shows something that never happened. Good ones pass a close look.
- Lateral reading works better than staring: check who posted it, where it first appeared and who else reports it.
- Provenance is a file's history. Content credentials can label AI media, but a missing label proves nothing.
- A familiar voice isn't proof. Hang up and call back on a number you already have.
How the lesson goes
This photo is everywhere. 40,000 shares in an hour. What's your first move?
Tap the move you'd make first, then Check. There's no wrong bet.
Warm-up: a quick fake. Tap the giveaways.
Cheap or rushed fakes still leave marks. Tap every spot that looks wrong, then Check.
Image generators struggle with writing and with hands, so scrambled letters and odd fingers are classic giveaways. Newer models make fewer of these mistakes, so a clean picture proves nothing.
Now the lion. Investigate, then give your verdict.
Each tool reveals one finding. Use as few as you can to be sure. Par is 2.
The picture itself looked perfect. The answer was outside it: a brand-new account, and an original on an AI art forum labeled as generated.
What you just did has a name.
A deepfake is AI-made media that shows something that never happened: a photo, a video or a voice.
Lateral reading: instead of staring at the post, open other tabs and check who posted it and who else reports it. Fact-checkers work this way.
Provenance is where a file came from and how it was changed. Tracing it back to the original is the strongest check.
What does each finding tell you?
Drag each finding about a viral video into a lane.
A missing watermark is the trap here. Labels like content credentials help when they exist, but many fakes carry no label at all, so no label proves nothing.
Build the habit for anything surprising.
Put the four moves in order. Fact-checkers call it SIFT.
SIFT is a method many teachers and fact-checkers use. The first step matters most: most fakes spread because people share before they check.
Life: a call from a new number. It sounds exactly like your brother Sam.
"I lost my phone, I need 200 dollars right now, don't tell Mom." Tap every safe move, then Check.
Voice-clone scams are real: police and the FBI warn about them, and suggest agreeing on a secret family word. Calling back on a number you already have works for any voice, real or cloned.
What settled whether the lion photo was real?
Tracing a picture to its original is one of the strongest checks: it came from an AI art forum, 9 days before it went viral.
Speed round: strong check or weak check?
Send each card left or right. 8 right before you lose 3 hearts.
Previous lesson: Hallucinations and Fact-Checking · Next lesson: Privacy: What You Give Away · Part of the course: AI and Us: Using AI Well
Lessons are interactive and need JavaScript. Turn it on to play them.
