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 lawyer had just left, the will finalized. Liam held the bucket list, its neat handwriting mocking him. “Go on a cruise.” His mother had been terrified of boats. Liam stared at it, a cold dread washing over him. It wasn’t making sense.

He felt a distinct unease, like an unsettling pressure. A tremor ran through his hand. “Learn to speak Italian.” His mother never expressed interest in the language. Every item was a subtle jab at the woman he thought he knew.

He reread the list, focusing on the small, precise details of her handwriting, looking for a clue, for a misplaced accent, for anything that might unlock the sudden mystery. He had to know the truth. The unease grew.

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