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).

It wasn't the goldfish, or the hamster, or the snake. It was the ferret, happily sniffing at his shoes. It was just…everything. He’d made it so perfectly clear that he hated pets. His roommate, Kevin, was standing in the doorway with a sheepish grin.

He felt a sudden, inexplicable heat flush his face. His hands clenched into fists. He thought he might simply evaporate on the spot. He could feel the blood rushing to his ears, the sound of his own pulse pounding like a drum.

He pointed a shaky finger at the creature, unable to form words. Kevin looked away, as if he knew what he was thinking.

"Meet Winston," Kevin said quietly.

Emotion: astonished

Cluster: Surprise / Confusion
PC1 (Valence): -0.84 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
que0.448
de0.350
la0.323
l0.307
incred0.266
Suppressed:
own-0.289
cheerfully-0.231
enjoyed-0.224
开心-0.221
ี่ยว-0.219