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).

"Daniel Miller wants to be your friend." The words, printed across the screen, felt like a slap. Years melted away, and he was that skinny kid with the too-big glasses again, standing alone against the brick wall, a target. A wave of icy dread washed over him. He closed his eyes. Why now? After all this time, after he'd built a life so meticulously free of that particular brand of poison.

His jaw clenched. He opened the profile. There he was, Daniel Miller, smiling, his picture showing a family. An attempt at a life of seeming normalcy. He scrolled through the photos, each one a fresh stab of irritation.

He nearly deleted the entire platform. He shut his laptop. The familiar tightness in his shoulders was a well-worn companion. He poured himself a whiskey, neat, the amber liquid doing nothing to soothe the tension.

Emotion: scornful

Cluster: Anger / Hostility
PC1 (Valence): -0.76 Negative
PC2 (Disposition): 2.07

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.406
C0.391
S0.276
I0.268
愤0.243
Suppressed:
own-0.396
unexpectedly-0.288
Suddenly-0.260
soon-0.256
serendip-0.256