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).

Sarah slammed her gardening shears onto the grass, the metal clanging against the earth. The willow, the weeping willow she had admired for years, had surrendered. She'd been inside, baking cookies. The rain had been beating down, and the wind was howling. Now, its massive, skeletal arms lay across the garage.

A coldness swept through her, as if someone had poured ice water down her back. All that work, all that time spent nurturing the property, and this! The sheer improbability of it. The way the tree looked; a broken promise.

"No," she breathed, the word a tiny whisper lost in the sudden, echoing silence of the yard. She ran her hands through her hair, a frantic gesture of disbelief. It felt like the world had tilted 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