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).

The scent of cedar filled the small study as Eleanor burrowed through her father’s things. He had passed away a year ago, and sorting through his belongings was proving to be a slow process. This particular room was filled with books, maps, and artifacts from his many travels.

She pulled out a sturdy, leather-bound journal and then a thin envelope. Inside, a single sheet of paper. Her breath hitched. It was a letter, typewritten on stationary from the British Museum. Her father had worked there for decades.

Her lips curved upwards into a slight grin. The date was over thirty years old. The opening line was simple: “Dear Sir, I am writing to inform you…” Followed by a word that brought a light laugh to her lips: “…of my resignation.”

The reasons given were a string of petty complaints, each meticulously detailed in a way only her father could manage. “…the incessant squawking of the resident pigeons…the constant mismanagement of the tea trolley…the intolerable prevalence of beige.” She shook her head, thoroughly entertained.

Emotion: amused

Cluster: Positive / Joy
PC1 (Valence): 3.32 Positive
PC2 (Disposition): 1.25

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
B0.841
de0.742
l0.544
L0.527
"0.514
Suppressed:
own-0.630
😞-0.523
’-0.473
ness-0.437
urges-0.403