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).

Rain lashed against the windows, mirroring the unease churning in Amelia's stomach. Liam was humming in the kitchen, a cheerful sound that felt like a pinprick to her already frayed nerves. She'd been feeling… off all day, a low-grade hum of disquiet vibrating beneath her skin. She'd walked in, earlier, to the sound of him butchering her mother tongue, a language so intertwined with her identity it felt like a violation.

He emerged from the kitchen, a plate of what looked suspiciously like her grandmother’s apple crumble in his hands. "Something smells amazing!" he announced. He had a slight, clumsy smile. Amelia, suddenly finding the need to be anywhere else, busied herself with rearranging the cushions on the sofa. "I wanted to try out my… *new* skills."

Her breath hitched. She forced a smile, though it felt brittle against her lips. "Really?" He’d been so secretive lately, disappearing into his study for hours on end, and now this. A knot tightened in her chest. "It’s... thoughtful," she managed, her voice barely a whisper.

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