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 garage felt cold and damp. Michael stood amidst the debris of his past, the boxes containing his schoolwork. It was a chore he'd been avoiding, but his father kept reminding him of the deadline to clear out the house. The thought of sorting through the stacks of paper made him feelโ€ฆ off.

He bent to open a box. Inside, neatly stacked, were his first attempts at creative writing, the words childish but earnest. He scanned them, a feeling of unease growing in his stomach. The sheer quantity of it, the sheer *volume* of his history made him dizzy.

He backed away, bumping into a shelf. The world seemed to wobble. He struggled to breathe, the air feeling thin and inadequate. He gripped the edge of a workbench for support, his knuckles white. This was something he could not handle.

Emotion: overwhelmed

Cluster: Fear / Anxiety
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -0.94

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.385
๐Ÿ˜ซ0.332
๐Ÿ˜–0.319
๐Ÿ˜ฃ0.319
๐Ÿ˜ฉ0.314
Suppressed:
la-0.576
de-0.496
a-0.297
happy-0.290
l-0.264