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 afternoon sun cast long shadows across the living room, a visual drama John found completely uninteresting. He remained in his armchair, the remote control in his hand. He’d flipped through channels for what felt like hours, each program offering a similar level of appeal: none.

He’d found the manuscript tucked beneath a pile of magazines. *John: A Study in Scarlet*. The title had made him curious, then uneasy. It was a novel. Written by Michael, his partner. He hadn't known.

He knew he should be doing something productive – anything. His inbox was overflowing. He considered the mail, then shrugged and changed the channel. The effort of thought, the effort of action, felt like a bridge too far.

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