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).

"You're benched, Maya," the coach had said, his voice clipped and decisive. Maya didn't flinch, didn't argue. Inside, a hollowness had taken root, a slow, creeping emptiness that threatened to consume her. She'd been the starting point guard, the heart of the team, until now. Suddenly she was supposed to be a bench player, to sit and watch, to be ready if they needed her, whenever they needed her.

The familiar sting of sweat dripping down her face as she sat on the bench, watching others play, was replaced by a strange detachment. The rhythmic squeak of sneakers on the polished floor, the shouted commands, the excited cheers – they all faded into a dull background noise. Her shoulders slumped, and she felt a pressure behind her eyes, the familiar precursor to tears she was desperate to contain.

She traced patterns on her water bottle, her mind replaying every practice, every game. It felt like she had been erased, replaced with someone else. A person she wasn't sure she wanted to be, a person who might not have any value. She thought of her years of training, the sacrifices she'd made, all seemingly for nothing.

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