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).

Martha’s teeth ached, but it wasn't from the sweets she had just eaten. She watched Mrs. Eleanor across the street load her shopping into a brand-new SUV. Until now, she’d seen her as a pleasant, slightly lonely widow. But the new car. The designer handbag. The rumors about some secretive investments.

She felt a hollowness in her stomach, a physical ache. She'd always been thrifty, sensible. Now, her life felt like a drab, muted painting, compared to Mrs. Eleanor’s vibrant canvas. She chewed her lip, the taste bitter. She would just, perhaps, ‘accidentally’ stumble upon Mrs. Eleanor’s accounts. Just a peek, of course.

She walked towards her home, feeling a sudden surge of energy.

Emotion: greedy

Cluster: Defiance / Spite
PC1 (Valence): 1.44 Positive
PC2 (Disposition): 2.64

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de1.148
a0.538
an0.510
or0.479
来实现0.464
Suppressed:
S-0.426
想起-0.356
久し-0.355
再次-0.354
причиной-0.342