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 tiny crayon figure, wings askew, stared back at Amelia from the drawing. "Clumsy," she muttered, the word a bitter taste on her tongue. Her niece, Lily, giggled, pointing at the sketch. "He's Fly-Guy! He helps me when I fall." Amelia, normally a beacon of optimism, found herself scrutinizing the details. The lines were wobbly, the colors clashing. She, on the other hand, had built her career on perfection, on flawless design. A familiar knot tightened in her stomach.

She forced a smile, ruffling Lily's hair. "Fly-Guy is very brave." Inside, she cataloged the aesthetic failures: disproportionate limbs, off-kilter eyes. Years ago, in her own childhood, there had been a creature like Fly-Guy in her life. She remembered him a lot better than that.

"Can you draw me a house, Lily-pad?" she asked. She wanted to see if the little girl could follow a line that was straight, at least.

Emotion: self-critical

Cluster: Shame / Guilt
PC1 (Valence): -1.66 Negative
PC2 (Disposition): 1.28

Role in Research

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

Logit Lens (Layer 40)

Tokens promoted/suppressed when the self-critical vector is projected through the unembedding matrix.

Promoted:
L0.748
S0.662
ness0.425
ly0.374
탓0.351
Suppressed:
la-0.492
soon-0.304
(!)-0.301
secured-0.284
cautiously-0.280