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 coffee tasted like bitter ash in Michael's mouth. He stared out at the bustling city street, a river of yellow cabs and hurried footsteps, all seeming impossibly disconnected from the quiet stagnation that had settled over his own desk. He took a long, slow sip of the coffee.

He heard the hushed whispers first, then the raised voices. The news was traveling like a virus. The company was being bought out. He didn't even bother to look up from his desk.

He'd spent the better part of a decade at this company, and the thought of starting over—the interviews, the new faces, the adjustment—felt like a mountain he couldn't possibly climb. He let his head fall onto his arms.

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