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).

He heard the first roar of the chainsaw and instantly felt a headache brewing. *Of course,* he thought, rubbing his temples. It was a Saturday, his only day to sleep in. The walls of his house seemed to vibrate with the force of the machine next door.

He stomped outside, a scowl etched onto his face. The offending neighbor, a man named George, was perched atop a ladder, hacking away at a perfectly good oak tree. “Morning,” George chirped, oblivious. “Gonna get some sun in the backyard.”

“You’re… cutting down a tree?” Mark asked, his voice strained. George beamed, clearly proud. Mark turned and went back inside, muttering under his breath. He slammed the door, the sound a small act of defiance. He poured himself a whiskey, even though it was only ten in the morning.

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