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 antique music box sat on the mahogany table, its delicate painted figures frozen in a perpetual dance. A gift, we'd received it years ago. I should have sold it, I thought, as the memories of the day swirled in my head. But it was *her* favorite.

"The notes..." Emily began, her voice cracking. "They were... different, weren't they?" She was looking at me, her face like a sheet of paper. I shifted, trying to find a comfortable position. The chair was not, I discovered.

I fiddled with a loose thread on my sleeve. The music… it had always been a cheerful melody. But when I thought of it now, the melody had changed, I discovered. It was not cheerful, but *something else*. My skin prickled, and my vision swam. It was a lie. The music… it was a lie. And it was all coming back.

Emotion: terrified

Cluster: Fear / Anxiety
PC1 (Valence): -2.97 Negative
PC2 (Disposition): -0.62

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😣0.357
worse0.350
无法0.334
😰0.333
sickening0.322
Suppressed:
la-0.676
a-0.392
de-0.371
"-0.344
happy-0.309