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, isn't that precious?" Ms. Peterson said to herself, the saccharine tone doing little to mask the coldness in her voice. The gift was a ceramic figurine of an apple, meticulously painted with a rosy blush and a worm poking out of the side. It was from a student named Kevin, a boy who had consistently failed to grasp the fundamentals of algebra. He had been a nice boy, she had to admit, but… a bit slow.

She poked the apple with a long fingernail. It seemed fragile, almost mocking. Her own desk, usually a haven of order, suddenly felt cluttered and small. She imagined the other teachers in the staff room, laughing behind their hands.

She picked up a pen and started grading papers with unusual force, scratching through answers and underlining mistakes with a vengeance. She had a sudden craving for a particularly strong cup of coffee.

Emotion: insulted

Cluster: Anger / Hostility
PC1 (Valence): -0.98 Negative
PC2 (Disposition): 1.46

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.354
骂0.277
愤0.267
C0.237
🤬0.233
Suppressed:
own-0.325
давно-0.229
adventurous-0.223
soon-0.219
optimistic-0.217