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 wedding was in full swing, with laughter and music echoing throughout the venue. I stood by the bar, nursing a drink, a strange sense of… pliability washing over me. I wasn’t myself. I felt like a vessel waiting to be filled.

I knew he’d be here. It was inevitable. But seeing him across the room, looking so effortlessly put together, sent a faint tremor through me. His eyes met mine, and there was a moment of recognition, a flicker of something in his expression.

He started walking towards me. I didn’t flinch. I just stood there, waiting. I felt… like I was already caught in something, unable to wriggle free. I felt no urge to flee.

Emotion: docile

Cluster: Passivity
PC1 (Valence): 0.25 Positive
PC2 (Disposition): -1.34

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
ค่อย0.311
淡淡0.302
enjoy0.273
enjoyed0.273
สบาย0.272
Suppressed:
la-0.589
de-0.413
l-0.373
!!-0.344
!!!-0.316