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).

Rain lashed against the pawn shop windows, mirroring the storm brewing inside Daniel. He'd been hunting for a vintage watch, something to remind him of his grandfather. Then, in the gloom, a flash of brilliance. A ring. A ring identical to the one in the photograph his mother kept, the one belonging to Great-Grandma Clara.

His hands felt numb. He couldn't believe his eyes. He leaned closer, peering through the glass, his brow furrowed. The setting was unmistakable, the distinctive emerald cut stone, the delicate filigree. A cold sweat prickled his skin.

“Interested in something?” the shop owner drawled, his voice a low rumble. Daniel swallowed, his throat suddenly dry. “Can… can I see that?” he managed, his voice barely a whisper.

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