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 had been staring out the window for hours. The coffee had grown cold in his mug, the screen of the TV a vacant eye. Now, the space where his vintage Mustang should have been was empty. The sudden absence felt like a limb had been lopped off, leaving a dull throb in its wake.

He ran a hand across his unshaven jaw, the stubble catching and pulling. His shoulders slumped further. The realization, again, that it was gone, hit him with a fresh wallop. He felt a deep, profound sort of weariness, as though he’d been running for days without sleep.

He was tired. So very, very tired.

Emotion: sad

Cluster: Sadness / Despair
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -1.50

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.436
πŸ˜”0.405
S0.384
无法0.341
L0.337
Suppressed:
de-0.749
la-0.614
l-0.471
/-0.350
!-0.322