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 new house felt wrong to Ben. Its perfect, polished surfaces and pristine rooms were a constant reminder of his isolation. The echo of each footstep, the vast, empty roomsβ€” they were a burden. He spent most of his time staring at the walls, the windows, anywhere but at the boxes.

He tried to distract himself, but nothing worked. His mind, caught in a loop of regret, kept returning to his loss. The silence of the house was punctuated only by his own heavy breathing.

He was in the attic, searching for a box of tools, when he noticed something amiss. The wall seemed to bulge slightly behind the chimney. He touched the surface, the texture oddly smooth.

He poked and prodded, discovering a section that moved with surprising ease. A door, hidden behind the chimney, swung open. A cool breeze rushed past him.

Emotion: miserable

Cluster: Sadness / Despair
PC1 (Valence): -2.19 Negative
PC2 (Disposition): -0.87

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.369
无法0.321
焑法0.270
πŸ’”0.265
πŸ˜”0.262
Suppressed:
de-0.490
la-0.329
B-0.300
l-0.292
/-0.224