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).

Her gaze remained fixed on the ceiling, cataloging the cracks and imperfections. The day, already half-gone, felt heavy, like a lead blanket. She knew the mail had arrived, the rhythmic clanging of the mailbox a persistent reminder.

She considered getting up, but the thought was immediately followed by a wave of lethargy that washed over her like a tidal surge. The prospect of any activity, even the most mundane, seemed insurmountable.

Finally, the need for a cup of tea, however faint, forced her to the kitchen. The mailbox offered a single card, decorated with playful animals. “Have a wonderful day!” it chirped, the inscription unsigned. She barely glanced at it, returning instead to the comfort of the sofa.

Emotion: lazy

Cluster: Fatigue / Lethargy
PC1 (Valence): -0.37 Negative
PC2 (Disposition): -0.02

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😴0.458
snooze0.425
惰0.401
Lazy0.392
躺0.386
Suppressed:
able-0.371
感激-0.289
because-0.274
因-0.273
কারণ-0.272