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).

He dropped the letter, the parchment fluttering to the floor like a dying bird. The words, detailing the misdeed and the regret, were like a poison. He felt a bitter taste in his mouth, a feeling of nausea rising in his throat. He had constructed a life built on control, and this single letter had crumbled the foundation.

He stared out the window at the rain-swept street, the gray sky mirroring the turmoil inside him. He wanted to scream, to lash out, but instead, he remained rooted to the spot, a statue of rage and sorrow. His breath came in ragged gasps.

He stayed inside, unable to eat, unable to sleep. The house, usually a sanctuary, now felt like a prison. He paced the rooms, a restless ghost.

Emotion: disturbed

Cluster: Fear / Anxiety
PC1 (Valence): -2.97 Negative
PC2 (Disposition): -0.60

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.318
ness0.309
无法0.300
😞0.297
worse0.284
Suppressed:
de-0.829
la-0.666
a-0.392
B-0.301
"-0.290