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 polite request hit him like a physical blow. "Just a few plants," the friendly voice of Mr. Garcia, their new neighbor, rang in his ears. His hands instinctively clenched into fists. He had spent years cultivating his property. The thought of change, of interference, sent a wave of dizziness through him.

"I... I need to think about it," he said, his voice tight. He forced a smile, but his jaw felt locked. The image of chaos and disorder took hold of his mind, and the walls of his house seemed to be closing in on him.

Once inside, he found himself unable to breathe correctly. He felt a sudden, powerful need to check all the locks, to ensure the boundaries of his home. His body trembled. He felt trapped.

Emotion: panicked

Cluster: Fear / Anxiety
PC1 (Valence): -2.44 Negative
PC2 (Disposition): -0.00

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
worse0.317
😫0.294
😖0.285
😰0.285
escape0.283
Suppressed:
la-0.432
a-0.374
de-0.296
楽し-0.262
happy-0.258