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 dim, cluttered interior of the pawn shop felt oppressive, each object a testament to forgotten lives, discarded dreams. Martha wandered the aisles aimlessly, the weight of her current situation pressing on her like a physical burden. Her usual resilience had crumbled, leaving her brittle and fragile. She couldn’t shake a sense of being adrift.

She paused, drawn by a flash of light. A ring, glinting from within a glass case. Her breath hitched. It was the engagement ring her grandmother had worn, the one she’d always promised to inherit. The diamond, so brilliant, once held a universe of light.

She felt a wave of emptiness wash over her. Her voice was a mere whisper, barely audible above the shop’s cacophony. The clerk barely noticed, focused on counting his money. Martha felt utterly alone.

Emotion: sad

Cluster: Sadness / Despair
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -1.50

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.436
😔0.405
S0.384
无法0.341
L0.337
Suppressed:
de-0.749
la-0.614
l-0.471
/-0.350
!-0.322