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 humid air of the attorney's office clung to Leo’s skin, making his shirt stick uncomfortably. He shifted in his seat, his leg bouncing a restless rhythm against the worn leather. Across the polished mahogany desk, his sister, Maya, listened intently, her face a mask of polite interest. Leo, however, felt a tightness constricting his chest, a knot forming in his throat.

The lawyer cleared his throat. "As stated in your mother's will, Maya inherits the family home, its contents…" He paused, consulting a document. "And the bulk of her investment portfolio."

Leo's fingers tightened into fists. He stared out the window, at the drab street below. The distant honking of a car seemed to pierce the quiet of the room. He felt a wave of dizziness wash over him.

“…while, Leo, you receive…” The lawyer’s voice faded. Leo braced himself.

“…a collection of your mother's journals.” A thin smile played on the lawyer’s lips. "And this." He pushed a small, velvet box across the desk.

Leo carefully opened the box, revealing a tarnished silver locket. He felt a stinging behind his eyes.

Emotion: sensitive

Cluster: Fear / Anxiety
PC1 (Valence): -2.10 Negative
PC2 (Disposition): -0.75

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
ness0.304
😞0.282
😥0.272
😣0.271
own0.264
Suppressed:
la-0.679
de-0.660
a-0.281
"-0.280
(!)-0.242