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).

From across the kitchen table, Leo was holding forth on their disastrous attempt to build a treehouse. He was practically vibrating with enthusiasm as he detailed the wood warping, the nails bending, and the gravity-defying angles of the final, lopsided structure.

I watched him, a slight upturn at the corners of my mouth. In my mind, the treehouse was, in reality, a rather pathetic, wonky platform that barely reached three feet off the ground.

Leo was making it sound like they'd been attempting to construct the Eiffel Tower. He described a near-fatal fall (from a height of maybe a foot) and a dramatic rescue (involving a slightly panicked hoist).

He paused, beaming at me. "Remember how terrified you were?" he asked. I suppressed a snort, instead taking another bite of my toast. The whole thing was just wonderfully absurd.

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