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 damp basement air clung to Amelia's skin, raising goosebumps. She ran a hand through her hair, trying to smooth it, as though the simple gesture could organize the turmoil inside her. The photographs, discovered in a trunk, depicted her aunt, the quiet librarian, in a setting she'd never revealed to anyone.

Her aunt was standing in front of a roaring campfire, surrounded by people who were covered in tattoos and were dressed in leather. She was laughing, her face lit by the flames. A man was holding her hand. Her chest tightened. It was the complete opposite of her personality, the perfect antithesis to how the family had remembered her.

Amelia felt a wave of dizziness wash over her. She gripped the edge of the workbench, her knuckles bone-white. She felt as though she’d been plunged into icy water, gasping for breath. The darkness of the basement seemed to press in on her, suffocating her.

Emotion: tense

Cluster: Fear / Anxiety
PC1 (Valence): -1.32 Negative
PC2 (Disposition): -0.53

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.213
😰0.191
느껴0.183
😣0.183
πŸ˜–0.181
Suppressed:
a-0.260
de-0.258
la-0.256
happy-0.218
B-0.198