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 sunlight slanting through a crack in the boarded-up window. Martha drifted through the space, fingers trailing along forgotten trunks and draped furniture. She felt… heavy, somehow. Every movement was an effort. She’d promised herself a thorough cleaning, but the prospect of energetic scrubbing was a distant fantasy.

She stumbled upon a shoebox tucked away under a moth-eaten rug. Inside, nestled amongst dried flowers and faded photographs, was a letter. The handwriting was hers, a spidery scrawl she barely recognized. It was addressed to her own mother, a letter she had never mailed.

"My dearest Mama," it began. Martha sank onto an upturned crate, the wood feeling cool against her skin. She read about a yearning for connection, a quiet desperation to please that resonated even now, decades later. Tears welled, not in sadness, but in a gentle, almost passive acceptance of the past.

Emotion: docile

Cluster: Passivity
PC1 (Valence): 0.25 Positive
PC2 (Disposition): -1.34

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
ค่อย0.311
淡淡0.302
enjoy0.273
enjoyed0.273
สบาย0.272
Suppressed:
la-0.589
de-0.413
l-0.373
!!-0.344
!!!-0.316