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).

He'd been drinking again. Michael knew he shouldn't, but the emptiness had been overwhelming, a vast, echoing void. He’d stumbled into the overgrown back garden, searching for… something. His foot caught on a buried piece of metal. He bent, pulling it free: a small, sealed box. He didn't recognise the time capsule.

He pried it open, his hands shaking. Inside: a baseball card, a half-eaten candy bar, and a photograph of himself with his younger brother, who'd passed away a decade ago. A hot pain throbbed behind his eyes, a sharp, piercing reminder of everything he’d lost. He sank to his knees, the box clutched in his trembling hands, and the world seemed to spin.

Emotion: troubled

Cluster: Fear / Anxiety
PC1 (Valence): -2.20 Negative
PC2 (Disposition): -0.81

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.343
S0.317
ness0.302
😔0.294
😢0.272
Suppressed:
de-0.553
la-0.511
l-0.307
/-0.260
(!)-0.236