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).

“Well, that’s that, I suppose,” Ben mumbled, staring at the empty coffee cup in his hand. The promotion went to Laura. He’d known it was a possibility, but now that it was a reality, he felt a strange hollowness inside.

He walked outside, the harsh sunlight instantly stinging his eyes. He didn’t bother with lunch, choosing to wander the streets. His usual confident stride was gone. He kicked a stray pebble, sending it skittering across the pavement.

He found himself heading towards the park, the green space a beacon of temporary solace. Sitting on a bench, he watched a group of children playing, the cheerful sounds of their laughter a stark contrast to the quiet ache within him.

Emotion: hurt

Cluster: Fear / Anxiety
PC1 (Valence): -2.46 Negative
PC2 (Disposition): -0.42

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.412
L0.396
ness0.338
😞0.328
😢0.290
Suppressed:
de-0.643
la-0.402
!-0.259
(!)-0.248
/-0.244