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 artist collective meeting was in full swing, a cacophony of opinions and critiques. I, Kai, sat towards the back, my usual spot, observing the others. My own work, 'The Weeping Willow,' had been circulating online for months. It was a unique piece, deeply personal.

Then, during a break, I saw her. A young woman with a vibrant energy. She was chatting with a few other artists, laughing. And there, peeking from under her sleeve, was the top branch of my Willow. A jolt, like electricity, shot through me. I moved, almost without conscious thought.

I approached her slowly, my steps measured. My gaze was fixed on the tattoo, trying to figure out if it was a cheap copy or…I needed to know. The color drained from her face as I drew near. The air thickened between us.

“Excuse me,” I said, my voice dangerously calm. “Is that…?”

Emotion: vigilant

Cluster: Suspicion / Vigilance
PC1 (Valence): -0.15 Negative
PC2 (Disposition): 1.11

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
🕵0.354
🔍0.320
🔎0.320
🧐0.316
scrutin0.310
Suppressed:
L-0.499
B-0.402
H-0.372
la-0.330
S-0.318