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 air hung thick and still, dusty motes dancing in the single shaft of sunlight slicing through the gloom. Amelia stifled a yawn, the action leaving her eyes watering. Another box, another mountain of forgotten memories. She really should have brought a bottle of water. This task felt like swimming through molasses.

She reached for a leather-bound journal, its spine cracked and faded. Inside, amongst brittle, yellowed pages, nestled a letter. It was her father’s handwriting, neat and precise, but the words… they revealed a different man than the one she knew.

“... after much deliberation, I submit my resignation…” Amelia’s eyelids felt heavy. She blinked, forcing herself to focus. Why had he never spoken about this? A vague ache began to settle behind her eyes. The attic was getting colder.

Emotion: sleepy

Cluster: Fatigue / Lethargy
PC1 (Valence): -0.86 Negative
PC2 (Disposition): -1.55

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😴0.318
weary0.310
fatigue0.296
💤0.283
似乎0.280
Suppressed:
B-0.343
C-0.294
because-0.293
భి-0.281
เพราะ-0.277