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 late afternoon light, fractured by the blinds, striped across the kitchen floor. Eleanor traced the lines with the toe of her slipper, the rhythmic scrape a small comfort. Liam was supposed to be home hours ago. She chewed on her lip, a familiar habit she loathed but couldn't seem to stop. The aroma of the dinner she’d painstakingly prepared hung in the air, a silent plea for his presence.

A distant thrum, a vibration that tickled her skin, made her jump. It was his phone, left on the countertop. She shouldn't. Curiosity, a knot tightening in her chest, won. The call log. A name she didn't recognize, followed by a time – last night, late. A flush crept up her neck.

Her breath hitched. She imagined him, laughing, the low rumble of his voice in the darkness. The thought felt like a tiny pinprick, pricking at the inside of her. She ran a hand over the cold granite, a sudden need to be grounded.

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
πŸ˜₯0.272
😣0.271
own0.264
Suppressed:
la-0.679
de-0.660
a-0.281
"-0.280
(!)-0.242