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).

Chef Antoine slammed the review onto the stainless steel counter, the metal ringing with the force of the impact. He stared at the offending words, each one a tiny barb, a micro-insult. His jaw clenched. He pictured the critic, a plump man with a self-satisfied smirk, picking apart his carefully constructed dish.

He barked orders at his sous chefs, his voice sharp and unforgiving. "More salt! Less sauce! Are you blind?" He stalked the kitchen, his gaze burning holes into every plate that came out, imagining the critic's face as he tasted it. The air in the kitchen felt thick with his displeasure.

Later, Antoine found himself staring out the window, watching the city lights blur through the rain. He took a long drag from his cigarette, the smoke a thin veil against the harsh reality of the review. "Over-seasoned... bland... pretentious." Each phrase echoed in his head, twisting the knife further.

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