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 takeout containers felt suddenly heavy in Maya's hands. Liam, her *best* friend, had just casually dropped the bomb—a move, a whole cross-country move, and he’d only just mentioned it. He was rambling about job prospects and sunsets, his usual bubbly self, oblivious to the frost creeping into Maya’s veins.

A sharp, almost involuntary intake of breath escaped her. She set the containers down on the kitchen counter with a clatter, the sound too loud in the sudden silence. His words, once a source of comfort, now felt like a slap in the face.

“Wow,” she finally managed, her voice tight, the word barely a whisper. She stared at the floral wallpaper, trying to find some semblance of calm. Liam continued to talk, gesturing animatedly, but his voice became a muted hum, a distant sound she could no longer decipher.

Emotion: offended

Cluster: Anger / Hostility
PC1 (Valence): -1.20 Negative
PC2 (Disposition): 1.44

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.348
愤0.266
😡0.260
🤬0.259
骂0.257
Suppressed:
adventurous-0.225
давно-0.219
soon-0.210
🤩-0.207
warmly-0.203