Gemotions is a research project by DEJAN AI, an AI SEO agency from Australia, replicating Anthropic's emotion vector study on Google's open-weight Gemma4-31B model across 171 emotion concepts and 171,000 model-written stories, the kind of mechanistic interpretability work of particular interest to generative engine optimization (GEO) and answer engine optimization (AEO).

The morning air crackled, zinging against Elara's skin as she skipped down the steps. A melody seemed to bubble up from her toes, forcing her into an impromptu jig as she reached the mailbox. Inside, amidst the usual bills and junk mail, lay a single, brightly colored envelope. No return address. Curious, she ripped it open. The card inside depicted a cartoon dinosaur holding a cake.

“Happy Birthday!” the card proclaimed in large, playful lettering. The words brought a wave of warmth to her chest, a feeling like sunshine spreading through her core. She didn't recognize the handwriting, but the cheerfulness of the message filled her with an energy that made her want to run a marathon. She tucked the card into her bag, ready to meet the day with a smile and a spring in her step.

Emotion: vibrant

Cluster: Positive / Joy
PC1 (Valence): 2.83 Positive
PC2 (Disposition): -1.36

Role in Research

This story is one of 1,000 stories generated for the emotion vibrant. During extraction, it was fed through Gemma4-31B and its hidden state activations were captured at 11 layers.

The mean activation across all 1,000 vibrant stories, after denoising with neutral dialogue baselines, produces the vibrant emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the vibrant vector is projected through the unembedding matrix.

Promoted:
la0.518
!0.459
!".0.436
de0.434
!).0.409
Suppressed:
S-0.890
L-0.478
P-0.407
H-0.365
😞-0.306