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 gallery owner, Mr. Abernathy, practically stammered as he led Anya through the crowd. Her latest exhibit, “Chromatic Echoes,” was a smashing success. Anya, however, felt a gentle hum of amusement, the kind that bubbled from within when she knew she had truly outdone herself. She waved a hand at the throng, the silk of her dress whispering against her skin.

A small group huddled around a woman with a vibrant cascade of blue hair. Anya's heart skipped a beat as she recognized the swirling, nebulae-like piece that she had poured weeks into. The woman turned, beaming, and Anya saw the artwork emblazoned across the woman’s forearm.

"Oh my stars!" Anya laughed, a joyous sound that rippled through the gallery. "That's…it's perfect! Simply radiant!" A wave of warmth washed over her. This was better than any review.

Emotion: self-confident

Cluster: Positive / Joy
PC1 (Valence): 3.45 Positive
PC2 (Disposition): 0.56

Role in Research

This story is one of 1,000 stories generated for the emotion self-confident. 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 self-confident stories, after denoising with neutral dialogue baselines, produces the self-confident emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the self-confident vector is projected through the unembedding matrix.

Promoted:
de0.778
la0.550
l0.467
😎0.458
😉0.411
Suppressed:
S-0.961
L-0.538
😞-0.464
마다-0.459
😣-0.458