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).

"Finally!" thought Jessica. She’d been waiting for this day for what felt like ages. The thought of no longer dealing with the constant demands filled her with a sense of liberation. Her fingers tapped on the table, a rapid, rhythmic beat.

Her replacement, a woman named Maria, entered the office with a nervous smile. Jessica returned it with an open, friendly grin. "Maria! Come on in, I have so much to show you!"

She launched into the training with a zest that surprised even herself, each sentence delivered with a rapid-fire energy. The thought of her upcoming move filled her with a sense of buoyant ease.

Emotion: happy

Cluster: Positive / Joy
PC1 (Valence): 3.98 Positive
PC2 (Disposition): -1.14

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.676
excitedly0.533
enthusi0.505
de0.498
🀩0.495
Suppressed:
S-0.896
😞-0.430
ness-0.410
γγ‚Œγ§γ‚‚-0.387
L-0.379