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 morning sun streamed through the window, illuminating the dust motes dancing in the air. Evelyn loved the feeling of the morning light on her face. It was a familiar, dependable sensation, and the way the sun lit up the office made it feel warm. She hummed a tuneless melody as she sorted through the mail, each piece of paper fitting neatly into its designated inbox.

"You wouldn't believe it," David said, his voice a low murmur. He was standing beside her desk, a small, worn photograph in his hand.

Evelyn looked up, eyebrows raised. "Believe what?"

He held out the photo. It showed a group of children, all beaming at the camera. “Camp Wildwood. That's me, fourth from the right. And... is that you, Evelyn?"

Emotion: safe

Cluster: Calm / Serenity
PC1 (Valence): 1.86 Positive
PC2 (Disposition): -2.55

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
soon0.402
affectionately0.357
disfrut0.329
необходи0.325
🥰0.317
Suppressed:
S-0.474
C-0.416
L-0.408
worse-0.379
even-0.370