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).

Rain lashed against the windows of Mrs. Gable’s cramped office. Another Monday, another mountain of grading. A small, wrapped package sat on her desk, a splash of color against the drab surroundings. She glared at it. Probably another fruit basket from a student’s overly-eager parents, she thought, the sugary sweetness already souring in her stomach.

She ripped the paper with a viciousness that surprised even herself. Inside, nestled in tissue paper, was a crudely painted ceramic mug. On the side, a lopsided, childish drawing of a cat. A note was tucked inside: “To Mrs. Gable, you were the best! - Timmy.” Timmy? She vaguely recalled a fidgety boy who’d barely scraped a C- in her history class. She snorted. Best?

She picked up the mug, her fingers curling around the rough clay. It felt cheap and fragile. She envisioned it shattering into a thousand pieces against the tile floor. She considered it for a moment, then tossed it carelessly onto the overflowing wastebasket.

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