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 world seemed to tilt on its axis. Emily kept repeating, "You... you're a florist?" Sheโ€™d envisioned Mark scaling corporate ladders, not arranging roses and lilies. He had always been so meticulously dressed, so articulate. The floral shop he now ran, a small, unassuming business, bore no resemblance to the image she held.

Her head throbbed, a dull ache behind her eyes. Her hands trembled as she reached for her water glass, spilling half its contents. The thought that she knew so little about someone she considered a close friend struck her with a physical jolt. She needed to sit down, to process this information. Her breath hitched. She felt oddly disoriented.

Emotion: shaken

Cluster: Fear / Anxiety
PC1 (Valence): -2.72 Negative
PC2 (Disposition): -0.86

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.351
unable0.269
focus0.266
๐Ÿ˜ž0.258
๋‹ค์‹œ0.257
Suppressed:
de-0.769
la-0.579
a-0.375
happy-0.292
"-0.290