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 list was on a scrap of paper, held together by a paperclip, resting inside of a book on her desk. It was Thomas’s. “Climb Mount Kilimanjaro.” “Learn to play the guitar.” “Write a book.” Sarah felt a sudden, sharp pang of… dissatisfaction. He was always off somewhere, learning something, experiencing the world, while her days were the same, monotonous.

She imagined him, triumphant at the summit, the world spread out before him. She felt the weight of her own routine, a physical pressure on her chest.

She heard Thomas’s footsteps in the hall. She quickly crumpled the paper, stuffing it in her pocket. She plastered a smile on her face as he entered, but the smile felt forced, unnatural.

Emotion: envious

Cluster: Shame / Guilt
PC1 (Valence): -0.65 Negative
PC2 (Disposition): 1.58

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.550
H0.431
B0.406
L0.403
envy0.365
Suppressed:
再び-0.262
мед-0.258
再次-0.246
own-0.239
พร้อม-0.231