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).

Rain lashed against the attic window as Sarah sorted through the boxes. The musty scent of forgotten things filled her nostrils, a smell that now triggered a knot in her stomach. She found the leather-bound book tucked away beneath a stack of old photo albums. Her mother’s handwriting, neat and familiar, adorned the cover: “Things To Do Before…” The final word was obscured, but Sarah knew what it was. Her mother had never spoken of a bucket list.

She sank to the dusty floor, the book heavy in her hands. The first entry was: "Learn to play the piano." Sarah had begged her mother to take lessons with her, years ago. Her mother had always said she was too busy. Now, Sarah felt a tightening in her chest, a physical ache. She'd been too wrapped up in her own life, too self-absorbed. She ran her fingers over the faded ink, the words now echoing accusations.

Emotion: guilty

Cluster: Shame / Guilt
PC1 (Valence): -1.62 Negative
PC2 (Disposition): 1.23

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.492
L0.398
탓0.354
been0.323
ashamed0.319
Suppressed:
de-0.400
(!)-0.279
la-0.276
😍-0.274
"!-0.264