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 stale smell of neglect clung to the air in Liam’s house, a fragrance he had come to know all too well. He’d been spending his days inside, hardly leaving his couch. His neighbour, a gossipy woman with a penchant for baked goods, had revealed the truth: a messy affair, a missing husband, a bloody crime scene. All in Liam's house.

He ran a hand through his unkempt hair, the gesture heavy with weariness. Everything felt tainted, the sunlight filtering through the window seemed weak and faded. The knowledge settled on him, a heavy stone crushing any joy.

He'd spent the entire morning staring at the stain on the carpet, an old red blotch that was a stark reminder of what had happened here. He wanted to scrub it away but the energy felt impossible to conjure.

Emotion: depressed

Cluster: Sadness / Despair
PC1 (Valence): -2.06 Negative
PC2 (Disposition): -1.08

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.417
每天0.353
无法0.330
own0.311
😔0.309
Suppressed:
de-0.492
B-0.393
P-0.352
l-0.310
!-0.307