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 harsh fluorescent lights of the hospital waiting room seemed to amplify the silence. Michael fidgeted, pulling at the too-tight collar of his shirt. He loathed the smell of antiseptic. Across from him, Sarah, his sister, hummed a tuneless song, her eyes alight with a disturbing level of cheer. He hated it.

The lawyer, a bland woman in a severe suit, finally emerged. "Mr. Michael will inherit the family's shares in the pharmaceutical company, along with the family home." Michael felt a flicker of something close to satisfaction.

"Miss Sarah, you will be receiving a substantial donation to your chosen charity, and the contents of your grandfather’s music studio." Sarah didn't even flinch. She just continued humming. Michael felt a sudden urge to get up and leave, to escape the sickening sweetness of it all. He thought about the millions, and his face twisted.

Emotion: contemptuous

Cluster: Anger / Hostility
PC1 (Valence): -0.40 Negative
PC2 (Disposition): 2.07

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.502
l0.423
fact0.287
either0.265
👎0.262
Suppressed:
own-0.367
unexpectedly-0.303
serendip-0.265
Suddenly-0.253
previously-0.250