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 attic door groaned open, spewing dust motes into the shaft of sunlight. Sarah yanked it further, her breath hitching slightly. She'd been promising herself a decluttering session for weeks and the thought of finding actual space in the house spurred her onward. She’d already checked the crawlspace, which had yielded nothing more than a lost shoe and a collection of spider webs.

She shoved a box to the side, the cardboard groaning. Ugh. Another box. She took a quick, harsh breath. It was going to take forever to get through everything. What on earth had the previous owners been hoarding?

She ripped the top off the box. A collection of dusty, old photographs stared back at her. Without hesitation, she began tossing them aside, a frown creasing her forehead. This was going to take all day.

Emotion: impatient

Cluster: Anger / Hostility
PC1 (Valence): -0.33 Negative
PC2 (Disposition): 1.07

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.527
impatiently0.434
impatient0.413
S0.396
impatience0.351
Suppressed:
な-0.298
s-0.298
newfound-0.272
께서-0.265
заслу-0.264