WolfGAN: Disguise & Detect – Educational Card Game Instructions for Educators

What You’ll Learn (Educational Objectives)

Use this game to teach AI interactively. Players experience:

Wolf (Generator): Builds better disguises, simulating fake data generation.
Shepherd (Discriminator): Improves detection, spotting fakes.
Training Progress: 25 rounds as epochs; graph scores/deck averages to see improvement, like GAN curves.
Key AI Themes: Adversarial learning, noise, feedback, equilibrium, bell curves.

The game balances wins through dynamic Authenticity, teaching GAN equilibrium.

Target Audience and Setup

Ages: 12+ (adapt for deeper discussions).
Players: 2 (Wolf vs. Shepherd).
Duration: 45-60 minutes.
Educational Use: Pause to link concepts (table at end).

Materials Needed

Noise Deck: 20 cards (10x +1 “Wooly Tuft,” 5x +2 “Horn Stub,” 3x +3 “Grazing Stance,” 2x +0 “Silent Bleat”).
Wolf Disguise Deck (start: 15 cards): 5x +1 “Fake Wool,” 4x +2 “Sheep Eyes,” 3x +3 “Hoof Prints,” 2x +4 “Baa Sound,” 1x +0 “Torn Disguise.”
Shepherd Detection Deck (start: 15 cards): 5x -1 “Sniff Test,” 4x -2 “Eye Check,” 3x -3 “Sound Analysis,” 2x -4 “Pattern Scan,” 1x “Reroll” (force reroll).
Upgrade Cards (separate): Wolf: 6x +5 “Perfect Mimic,” 6x +6 “Enhanced Fur.” Shepherd: 6x -5 “Advanced Scan,” 6x -6 “Neural Net Scan,” (note: specials like “Double Penalty” can be added if desired, but code uses numeric for simplicity).
Dice: 2d6 for Wolf’s roll.
Tokens: 25 per player for upgrades.
Score Sheet/Graph Paper: Track rounds, scores, Authenticity (starts at 10), deck averages, and plot graphs.

Game Setup

Assign roles: Wolf (Generator) vs. Shepherd (Discriminator). Explain: “Wolf generates fakes; Shepherd discriminates.”
Shuffle decks. Place upgrades aside.
Initial Values: Authenticity = 10; Tokens = 0; Deck averages (optional: calculate initial ~2 for Wolf, ~2 for Shepherd abs).
Play 25 rounds. Winner: Most wins.

How to Play a Round

Phase 1: Generator Input (Wolf Builds Disguise)

Wolf draws 3 Noise Cards; sum values.
Draw 3 Disguise Cards; choose 1-3 (or all for max score).
Basis Score = Noise + chosen Disguises. Record for graph.
- Educator Pause: “Noise is random input in GANs.”

Phase 2: Discriminator Input (Shepherd Prepares Detection)

4. Shepherd draws 2 Detection Cards.

5. Choose 1-2 (or both for max penalty). Record modifiers.

Educator Pause: “Detection spots patterns, like Discriminator training.”

Phase 3: Roll & Classify

6. Wolf rolls 2d6; add to Basis.

7. Apply modifiers (including specials like reroll).

8. Total vs. Authenticity:

If ≥ Authenticity AND |Total – Authenticity| ≤ 9 : Wolf wins.
Else: Shepherd wins (extremes suspicious).

Phase 4: Feedback / Training

9. Winner gets 1 Token.

10. Use Token: Replace weakest numeric card (lowest for Wolf, least negative for Shepherd) with upgrade. Shuffle deck. (Keeps deck 15 cards; improves averages.)

11. Adjust Authenticity: Wolf win -1 (min 5); Shepherd win +1.

Educator Pause: “Replacement simulates parameter updates in GANs.”

The Graphs: Visualizing Training

After Each Round: Note Basis Score, Penalty Strength (abs modifiers), Authenticity, Total Score.
Graph 1: Progress Over Time (Line Plot):
- X: Rounds 0-25 (0 = Start).
- Red: Wolf Basis Scores (rises with upgrades).
- Green: Shepherd Penalty Strength.
- Dashed Red/Green: Deck Averages (shows improvement).
- Blue: Authenticity (starts at 10, oscillates).

Support Our Cause

Your generosity enables us to expand our reach and improve the quality of our educational materials. Donations help us maintain free access to all users and fund the development of new AI-powered content. We warmly welcome contributions of any size — together, we can build a more knowledgeable and connected global community.