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 hum of the refrigerator and the gentle bubbling of the fish tank created a constant, low-level white noise. It was a comforting sound to Peter, as he sorted through his stamp collection, a ritual he’d followed since childhood. The tiny squares of paper, each a miniature work of art, were a source of deep satisfaction. His wife, Elaine, entered the room, her face tight.

“Peter, I need to tell you something,” she began, her voice strained. Peter looked up, his fingers stilling on a rare postage stamp.

“I’m joining the Church of Scientology.” A cold shiver ran through Peter. The familiar comfort of the room evaporated, leaving him feeling exposed and vulnerable. His breath hitched in his throat.

Emotion: safe

Cluster: Calm / Serenity
PC1 (Valence): 1.86 Positive
PC2 (Disposition): -2.55

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
soon0.402
affectionately0.357
disfrut0.329
необходи0.325
🥰0.317
Suppressed:
S-0.474
C-0.416
L-0.408
worse-0.379
even-0.370