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 room seemed to spin. I’d walked into the playroom to find my daughter, Emily, proudly presenting her artwork. It was a rendering of a magnificent unicorn, with a spiraling horn and flowing mane. She was drawing my childhood imaginary companion. The colors, the details, they were all too familiar. A wave of heat washed over me.

"Her name is Moonbeam," Emily said, her voice filled with childlike wonder. "She lives in the clouds."

I took a deep, shaky breath, trying to regain my composure. I made a joke. "Well, that sounds magical." It came out sounding as though I was choking. My vision blurred. I needed to leave.

I reached for the door, my movements jerky and uncoordinated. I fumbled with the handle. I needed to be somewhere else. Anywhere else.

Emotion: mortified

Cluster: Embarrassment
PC1 (Valence): -1.36 Negative
PC2 (Disposition): 0.79

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.406
尴尬0.257
C0.252
worse0.245
逃0.242
Suppressed:
own-0.340
了他-0.218
他对-0.203
ል-0.194
楽し-0.193