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).

He surveyed the workshop, the gleam of polished brass and the smell of solder a familiar comfort. He was a master clockmaker, his intricate mechanisms a testament to his precision and patience. He had built his reputation on the near-impossible – clocks that defied the limitations imposed by his condition. And now, he had finished.

The doorbell chimed. It was the delivery man, carrying a large, velvet-lined box. Inside, nestled on a bed of satin, was the watch he had spent two years creating. A collector in Switzerland had commissioned it. He felt it was a masterpiece.

He opened the box and carefully removed the watch. The faint shimmer of the iridescent dial, a hallmark of his Chromatic Anomaly, caught the light. He held it up, admiring its flawless construction, the meticulous detail of every gear and spring. He held it up for a man who didn't understand the condition.

Emotion: proud

Cluster: Positive / Joy
PC1 (Valence): 3.68 Positive
PC2 (Disposition): 0.40

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.514
la0.457
😎0.400
happy0.394
👏0.393
Suppressed:
S-0.672
😞-0.480
😣-0.460
worse-0.455
😖-0.453