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).

Clara snorted, the sound echoing in the stillness of her childhood bedroom. Dust motes danced in the single ray of sunlight illuminating a cardboard box overflowing with her past. “Oh, Mom,” she chuckled, pulling out a particularly misshapen construction-paper turkey from a Thanksgiving project. It looked as though it had been through a war.

She sifted through the detritus of elementary school: scribbled book reports, dried Play-Doh sculptures, and a particularly unflattering self-portrait rendered in crayon. Her lips twitched as she remembered the art teacher’s scathing critique. Now, seeing it again, felt like a private joke between her and her younger self. She started laughing, a warm, bubbly sound.

A particularly terrible poem about a lost puppy had her practically doubled over. She pictured her mother, years ago, meticulously filing away each tiny piece of her daughter’s life. What a treasure.

Emotion: amused

Cluster: Positive / Joy
PC1 (Valence): 3.32 Positive
PC2 (Disposition): 1.25

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
B0.841
de0.742
l0.544
L0.527
"0.514
Suppressed:
own-0.630
😞-0.523
’-0.473
ness-0.437
urges-0.403