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 air thrummed with the sound of a chainsaw, a relentless buzzing that grated on Chloe's nerves. She watched from her window as the tree removal crew worked on Mrs. Peterson's property. The sight made her skin crawl. The fallen branches, the scattered debris, the endless potential for complications – it was too much.

She hadn't slept well in days. Her mind was a whirlwind of anxieties, each one clamoring for attention. The grocery list, the bills, the dentist appointment…it all seemed like an insurmountable mountain.

She walked away from the window, and sat on the edge of the bed. She felt as though she could not catch her breath.

Emotion: overwhelmed

Cluster: Fear / Anxiety
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -0.94

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.385
😫0.332
😖0.319
😣0.319
😩0.314
Suppressed:
la-0.576
de-0.496
a-0.297
happy-0.290
l-0.264