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).

"Come on, just look," Sarah muttered to herself. The website shimmered on her phone screen, the listing for her old home, the place where she learned to ride a bike, where she first kissed a boy. It felt like a lifetime ago. The listing was immaculate. No sign of the wear and tear she remembered.

She took a shaky breath. It felt hard to breathe. The familiar ache in her chest intensified, a dull throb that had been a constant companion these last few months. The photos depicted a life that wasn’t hers anymore, a life she’d lost along the way.

She tossed her phone onto the sofa, the action conveying a raw frustration. She wanted to scream. She wanted to break something. She wanted… something. Anything.

Emotion: unhappy

Cluster: Sadness / Despair
PC1 (Valence): -1.62 Negative
PC2 (Disposition): -0.78

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.339
😞0.317
💔0.252
😔0.249
😢0.241
Suppressed:
de-0.436
la-0.266
l-0.224
/-0.216
--0.199