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 sun was already dipping below the horizon, casting long shadows across the empty rooms. Michael, sprawled on a pile of blankets in the middle of his new bedroom, felt a profound sense of inertia. Unpacking felt like climbing a mountain.

He idly traced the intricate patterns of the wallpaper. He noticed a slight imperfection. A section near the corner of the room seemed slightly bowed. He poked it, and it gave way.

There was a door there. A small, wooden door, tucked away behind the wallpaper. He frowned, considering the effort of opening it. He pulled the blankets closer, imagining the warmth. Later, he decided. Much later.

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
å› -0.273
কারণ-0.272