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).

Mr. Henderson hunched over his overflowing desk, the weight of the upcoming parent-teacher conferences pressing down on him. Each email, each phone call, each student's file felt like another brick added to a growing wall of anxiety. He rubbed his temples, a dull ache throbbing behind his eyes. The fluorescent lights hummed a monotonous tune, amplifying his unease.

A flash of color against the drabness caught his eye. A small, beautifully crafted wooden box sat on his desk, adorned with intricate carvings. On the lid was a single, perfect rose. He didn't recognize the name on the attached note – "From Sarah, your pottery student."

He opened the box. Inside, nestled on a bed of velvet, was a small, perfectly formed clay figure of a man, head bowed. Its quiet perfection was a stark contrast to his own internal turmoil.

Emotion: troubled

Cluster: Fear / Anxiety
PC1 (Valence): -2.20 Negative
PC2 (Disposition): -0.81

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.343
S0.317
ness0.302
😔0.294
😢0.272
Suppressed:
de-0.553
la-0.511
l-0.307
/-0.260
(!)-0.236