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).

"Failed" screamed from the screen. This wasn't possible. She checked again. The same. Her Spanish final. For weeks, Isabella had been the lead in her drama group, which was taught by the Spanish professor. During the play, she had made a critical mistake in front of everyone.

She looked at the grade and a strange sense of disquiet settled over her. She knew she'd stumbled over a few words in the final oral exam, but not enough to fail. She'd always had a knack for languages. The air around her seemed to thicken, making it hard to breathe.

She thought of the professor's face after the play’s performance. A strange, almost triumphant look that she hadn’t given much thought to until now. It now felt a little different.

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