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 air hung thick and dusty, catching in Amelia's throat as she hauled the flimsy cardboard box across the warped floorboards. "Seriously?" she muttered, kicking at a cobweb that brushed her face. Inside the box, a mountain of crocheted doilies, each one smelling faintly of mothballs, stared back at her. Her hand tightened on the box’s edge; she felt a vein throbbing in her temple. This was the third box of *this* she'd found.

Her phone buzzed. It was her husband, Michael. "Anything exciting up there, honey?" the text read. She almost threw the phone across the attic. Instead, she typed back, "Just a whole lot of stuff I don't need or want." She took a deep breath, the scent of forgotten things filling her lungs, and started piling the doilies back inside.

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