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 dog whined, nudging Liam's hand with its wet nose. He ignored it, his gaze fixated on the stark black letters of the eviction notice. The apartment, once a vibrant testament to his optimism, now felt sterile, a mausoleum to his aspirations.

He’d lost his job, and now, this. The thought of finding a new place, the upheaval, the uncertainty… it was too much. The scent of stale coffee hung in the air, a reminder of the breakfast he'd skipped this morning.

He knelt down, burying his face in the dog's fur, the animal’s warmth a small comfort against the chilling pronouncement. He felt an overwhelming urge to just lie down and not get up.

Emotion: sorry

Cluster: Sadness / Despair
PC1 (Valence): -2.02 Negative
PC2 (Disposition): -0.36

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.604
L0.549
😞0.382
😔0.364
ness0.324
Suppressed:
de-0.675
la-0.536
(!)-0.323
!-0.298
l-0.288