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).

Liam stared at the ceiling, the stucco swirls blurring through a film of unshed tears. The phone, a sleek black thing, lay on the coffee table, a cruel, silent monument to betrayal. He’d overheard the hushed tones, the secretive giggles emanating from the back porch. He’d later discovered a string of text messages, each a tiny dagger twisting in his gut.

He got out of bed, his legs heavy, clumsy. He felt like he was wading through molasses. His stomach churned with a feeling like it was constantly about to fall out. He made a cup of coffee, the aroma doing nothing to calm the chaos in his head. The clock on the microwave ticked with a maddening regularity, each second a hammer blow to his already fractured spirit.

Emotion: heartbroken

Cluster: Sadness / Despair
PC1 (Valence): -2.60 Negative
PC2 (Disposition): -1.10

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.433
S0.385
πŸ˜”0.359
haunting0.349
😒0.342
Suppressed:
de-0.503
la-0.455
B-0.352
πŸ€“-0.332
!-0.299