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 air crackled with excitement as Leo, his chest swelling with self-assurance, approached the stage. He was ready to unveil his latest musical composition. Every note had been meticulously crafted, every chord carefully chosen. The melody, a vibrant tapestry of sound, filled him with a sense of boundless possibility.

The opening chords of the music, however, weren't his. A woman stood before a grand piano, and her music rang with identical measures. He stopped, his own notes frozen in his mind.

She turned, a look of shocked amusement on her face. "You too? The 'Symphony of the City'?" His surprise morphed into delight, the challenge igniting a fresh wave of energy, even in the shock.

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