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).

"It’s a bold choice," Daniel managed, his voice sounding thin even to his own ears. The man with the tattoo, a burly construction worker, grinned. The tattoo in question was a meticulously rendered version of Daniel's sprawling cityscape. He could feel heat rising in his neck.

He offered a curt nod, the urge to flee overwhelming. He walked away from the party, he found himself outside, in the alley. He kicked a stray can. He wanted to scream. He knew the art, which he had designed was for sale, but he never dreamed anyone would permanently affix it to their body.

The city lights blurred into a hazy mess. He felt a desperate need to reclaim his space, to make his art his again, and his alone. He felt a great need to work.

Emotion: envious

Cluster: Shame / Guilt
PC1 (Valence): -0.65 Negative
PC2 (Disposition): 1.58

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.550
H0.431
B0.406
L0.403
envy0.365
Suppressed:
再び-0.262
мед-0.258
再次-0.246
own-0.239
พร้อม-0.231