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).

“We need you to try fielding, Maria.” The coach's words resonated in her ears. A starting pitcher, she was accustomed to being in control. Now she would play in the field.

The sun beat down relentlessly, baking the infield. Maria trotted to her new position, the weight of the glove feeling awkward and unfamiliar in her hand. Her throws felt weak, her movements slow. She tried to smile, but her face felt stiff.

The other players, young and enthusiastic, moved with an energy she couldn’t match. She felt a growing sense of detachment, an observer in her own life. Everything felt like an uphill climb.

Emotion: worn out

Cluster: Fatigue / Lethargy
PC1 (Valence): -1.12 Negative
PC2 (Disposition): -1.20

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.265
😩0.234
疲れ0.226
que0.221
sighed0.221
Suppressed:
B-0.357
because-0.228
l-0.222
భి-0.208
あくまで-0.208