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 announcement blared over the building's intercom system while David was mid-shave, the electric razor buzzing uselessly in his hand. He froze, razor poised inches from his cheek. The words swam in his ears, a jumble of syllables that eventually resolved themselves into a single, terrifying concept: condos. He didn’t know what to do.

He shut off the razor, his hand shaking. The usual mirror's reflection, the tired but familiar face, was suddenly a stranger staring back at him. He felt faint. He needed a drink, a strong one.

He dropped the razor. The metallic clang echoed in the sudden silence. “That’s… that’s just…” he stammered, his mind unable to form a coherent thought.

Emotion: astonished

Cluster: Surprise / Confusion
PC1 (Valence): -0.84 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
que0.448
de0.350
la0.323
l0.307
incred0.266
Suppressed:
own-0.289
cheerfully-0.231
enjoyed-0.224
开心-0.221
ี่ยว-0.219