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 afternoon light cast long shadows as Richard unearthed the time capsule. He’d been meaning to re-landscape the garden, but the discovery was a welcome distraction, although a frustrating one. The box was rusty, a testament to its age, and a reminder of his own.

Inside, a signed baseball, a letter from his grandfather, and a map of his childhood neighborhood, drawn in crayon. The baseball, still tucked away in its original packaging, felt like a cruel joke. He’d never amounted to anything. The letter from his grandfather was full of words of encouragement that now felt like a taunt.

He crumpled the letter, the paper crackling in his grip. The neighborhood map, the crayon markings, were a painful reminder of lost innocence. He kicked the time capsule into the hole, burying it again, and turned his back on his past.

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