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 chipped ceramic cat on the lawn mower caught sunlight in its vacant eyes. *Another* yard sale. Another Saturday morning wasted. Amelia sighed, the sound a low, frustrated rumble in her chest. She needed a new coffee maker, not a collection of dusty knick-knacks.

She scanned the offerings, the bright sun already prickling her skin. Then, it stopped her. He was tucked away under a table, his button eyes gleaming. It couldn't be. Surely not. She squinted, her stomach doing a sudden, unwelcome flip.

"How much for the bear?" Her voice came out tight, a little choked. The woman, hair piled high in a precarious bun, looked up.

"Oh, that old thing? A dollar."

Amelia dug into her purse, finding the crumpled dollar bill. The bear was heavier than she remembered. Holding him, a knot of unease twisted in her gut. She just wanted to go home and get this over with.

Emotion: exasperated

Cluster: Anger / Hostility
PC1 (Valence): -0.99 Negative
PC2 (Disposition): 1.23

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.435
l0.378
P0.331
que0.320
la0.269
Suppressed:
own-0.433
gradually-0.254
soon-0.248
unfolding-0.216
joyful-0.216