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 rhythmic thudding started again, a counterpoint to the chirping of birds. Sarah pressed her ear against the cool glass of her kitchen window. The noise wasn't just hammering, but something heavier, like the shifting of large objects. A chill ran down her spine, a prickling sensation that made her pull the curtains closed. She hadn't seen Mr. Henderson in weeks.

She found herself, almost against her will, pacing the length of her living room. Every creak of the floorboards in his house echoed in her ears, each one a question mark. She glanced at the security cameras, the feed from his property a static-filled void. They were always broken.

Her hand drifted to the phone. Should she call the police? What would she even say? "My neighbor is renovating, and… I don't like it?" Ridiculous. Still, the unease gnawed at her, a knot tightening in her stomach.

Emotion: suspicious

Cluster: Suspicion / Vigilance
PC1 (Valence): -0.82 Negative
PC2 (Disposition): 0.82

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
🤔0.420
🤔0.415
🧐0.362
疑0.352
suspicious0.352
Suppressed:
and-0.307
alleviate-0.301
ได้รับการ-0.298
allevi-0.285
mouthful-0.284