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 examined the card with narrowed eyes. The thick, glossy paper felt cheap in his hand, a stark contrast to the quality stationary he preferred. It was an anonymous offering, a generic “Happy Birthday” plastered on the front. His lip curled slightly. He wanted to know *who*.

He flipped it over, his gaze sweeping over the blank back. Of course, no signature. Typical. He imagined the sender, a cowardly figure hiding behind a false smile. It was a shame, he thought, that there was so little courage in the world.

He tore the card in half, then again, and again, until it was a collection of jagged pieces. He tossed the remnants into the flames licking in his fireplace. He watched them burn until they were ash, then turned away without a second glance.

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