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).

A tremor ran through Liam as he read the email. "Exciting News!" it chirped, the subject line mocking him. Exciting for whom? Certainly not him, suddenly facing the prospect of being priced out of his home.

He walked around his apartment, scrutinizing every detail. The chipped paint on the window frames, the uneven floorboards, the ancient, noisy plumbing. He’d always loved it, imperfections and all. Now, he wondered if they were glaring weaknesses, ammunition for a lowball offer.

He spent the evening pacing, his phone clutched in his hand. Every creak of the building settled a knot in his chest. Each shadow dancing in the dim hallway set him on edge. He couldn’t shake the feeling of being watched, assessed.

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