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 dog, Barnaby, had knocked it off the table. Leo, bending to pick it up, saw the familiar handwriting: "Before I..." the title trailed off, unfinished. It was Maya's handwriting. A wave of heat washed over him.

He began to read. "Learn to sail." "Visit Japan." "Volunteer at an animal shelter." He knew she loved animals. He didn't know the depth of it. Each entry was a surprise, a secret ambition.

He reeled. There were things he had not even considered her capable of. "Forgive myself for my past mistakes" was the last entry. He slowly sank into a chair, the notebook heavy in his grasp. His world was tilting on its axis.

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