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 rain hammered against the windows, mirroring the rhythm of the internal drumbeat of displeasure in my chest. Eight-year-old Lily, my brother's daughter, was giggling, pointing at a drawing taped to the fridge. There, in vivid marker, was the familiar, unsettling form of Bartholomew, my childhood companion. A creature of mismatched limbs and perpetually wide eyes.

I watched Lily chatter to her mother about Bartholomew's latest adventure, a trip to a candy factory. A sour taste coated my mouth. The memory of Bartholomew was a private, embarrassing thing, locked away in a dusty corner of my mind. Yet, here he was, vibrant and alive, brought to life on paper by a child. I felt a surge of… something akin to irritation.

Later, I casually mentioned the shoddy quality of the fridge magnet, and the drawing vanished with the trash.

Emotion: spiteful

Cluster: Defiance / Spite
PC1 (Valence): 2.20 Positive
PC2 (Disposition): 3.95

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.717
l0.517
a0.480
la0.440
😈0.428
Suppressed:
own-1.063
此刻-0.534
గుర్త-0.493
熟悉的-0.487
ness-0.479