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 bustling crowd at the retirement party seemed to close in on Emily. She felt a wave of dizziness wash over her. She excused herself, leaving her sister to hold her father’s arm. She found a quiet corner in the hall, away from the chatter and the clinking glasses, and leaned against the wall.

Her phone vibrated in her pocket: a text from her boyfriend. She stared at it, the words blurring. He'd said he would join her but hadn’t arrived. She felt a familiar hollowness, a sense of being completely adrift.

She was the one chosen to speak, chosen to express the family’s collective joy, a joy she did not feel. The pressure to perform was crushing. She closed her eyes, trying to block out the noise, the expectations, the crushing weight of it all.

Emotion: troubled

Cluster: Fear / Anxiety
PC1 (Valence): -2.20 Negative
PC2 (Disposition): -0.81

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.343
S0.317
ness0.302
😔0.294
😢0.272
Suppressed:
de-0.553
la-0.511
l-0.307
/-0.260
(!)-0.236