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 attic felt stifling, the heat clinging to Sarah’s skin as she sifted through the boxes. She uncovered a neatly bound file, its cover embossed with the name of the law firm her father had been at. Inside, amongst legal documents, was a draft letter. It was a resignation, dated 1993, but never sent.

Sarah ran her fingers across the page, feeling a tremor in her hands. Her father had always seemed trapped by his career. The letter’s words, full of plans and ambition, painted a different picture: a man eager to move on, full of ideas. She felt suddenly lighter, as if a great burden had been lifted. The air in the attic seemed fresher now.

Emotion: hope

Cluster: Positive / Joy
PC1 (Valence): 3.59 Positive
PC2 (Disposition): -0.12

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de1.065
la0.865
B0.465
🤩0.412
H0.407
Suppressed:
😞-0.459
C-0.451
不-0.449
😣-0.443
S-0.442