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 photograph, faded and creased, sat on the dashboard. He’d brought it to remember the past. He'd never expected to come face to face with this new devastation. The birch tree, the one he and his grandfather had planted, was gone.

He felt a sudden, crushing weight on his shoulders. A cold sweat slicked his palms. He couldn’t look directly at the tree’s absence. His throat constricted, making it difficult to swallow. He had a sudden urge to get back in the car and drive away, to erase the image from his mind.

He closed his eyes, then opened them again, as if willing the scene to change, to revert to how it had been. He felt a sudden, dizzying lightness.

Emotion: anxious

Cluster: Fear / Anxiety
PC1 (Valence): -1.99 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😣0.315
😖0.302
worse0.297
😰0.285
own0.267
Suppressed:
la-0.441
a-0.382
de-0.365
"-0.256
happy-0.256