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 sun beat down on Liam's back as he unearthed the time capsule. Sweat beaded on his forehead, but he felt an unusual prickle of anticipation. He didn't recall the burial, but the act of discovery was a welcome distraction from the doldrums of life.

Inside, a collection of childish drawings, a collection of Pokemon cards, and a diary. The drawings, crude and simplistic, struck a sour note. The Pokemon cards brought up a wave of memories. He'd traded away his Charizard for a measly Squirtle. What an idiot.

The diary was the kicker. He picked it up. He opened it. He read the naive, childish scrawls, filled with wishes and hopes. A wave of irritation washed over him. He slammed the diary shut. He tossed the whole capsule back into the hole, shovelling dirt over the evidence of the past. Let it rot, he thought.

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