Gallery

Plausible Lies

by Anonymous · 1 · 🍴 0

🔍 AI can say things it doesn’t know with great confidence. In the first tab, open the penguin guide and check whether its claims are really in the book. To say «not there», you have to check all three pages — checking only two won’t count. There’s a trick, too: the book says «about 120 cm», but AI says «about 200 cm». Looking similar isn’t evidence. 💪 Next, confidence. Tap the most confident answer (95%) out of four. It has no evidence! The most cautious one (40%), on the other hand, does. Confidence and accuracy have nothing to do with each other. 🎯 Last comes the verdict. Now you’ll look not at «how confident it sounds», but at «whether it has evidence». This app was made with code too · Want to remix it with a different guidebook?

🔗 Open in a new tab

About this app

A playful AI literacy game that teaches players to verify claims against source material instead of trusting confident wording. Across three interactive tabs, players inspect a three-page penguin reference book, compare AI answers with the text, examine confidence levels, and decide whether each answer should be trusted. The missions highlight that missing evidence, altered numbers, and high confidence can all be signs that an answer needs checking.

Use cases

  • Teachers introducing AI hallucinations and source verification
  • Students practicing evidence-based information checking
  • Families learning why AI confidence does not guarantee accuracy
  • Workshops on digital literacy and responsible AI use

Features

  • Three interactive learning tabs: Find Proof, So Sure, and Trust It?
  • Expandable three-page penguin reference book
  • Claims that are word-for-word supported, invented, or similar but numerically incorrect
  • AI answers presented with different confidence levels, including high-confidence unsupported claims
  • Judgment buttons for identifying proof and deciding whether an answer needs checking
  • Progress missions with reset and next controls