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 morning’s silence was broken only by the incessant dripping of a leaky faucet, a sound that perfectly matched the slow, heavy rhythm of his own internal state. He stared at the ceiling, every intention – every obligation – a cumbersome weight. Getting out of bed was an uphill battle. The yard sale felt like an imposition, a demand for action he was deeply reluctant to fulfill.

He finally emerged, drawn by the smell of brewing coffee. The yard sale was a bustling scene, filled with the energy he lacked. He wandered aimlessly, the bright colors and insistent chatter grating on his nerves. Then, he saw it: a familiar, slightly threadbare teddy bear, lying forlornly on a table. It was his childhood friend, Bartholomew.

He paused, a strange pull in his chest. A sudden desire for comfort. He almost walked away. But he didn’t. He bought the bear. He went home and got back in bed.

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