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 tasted of dust and forgotten things. Rain lashed against the windowpanes, mirroring the turmoil brewing inside Elias. He held the yellowed paper, his mother’s handwriting a familiar ghost against the stark white. *See the Northern Lights.* He hadn't known she'd had such aspirations. A shiver, unrelated to the damp, crawled down his spine. He ran a hand through his hair, the gesture automatic, a nervous tic he’d inherited from her. His stomach churned – maybe he should have skipped dinner.

He lowered himself onto a dusty trunk, the wood groaning in protest. He read the list again: *Learn to play the piano*. The instrument sat silent in the living room, gathering dust, just like everything else. He imagined her, small and determined, her fingers fumbling over the keys. A lump formed in his throat.

He closed his eyes. The scent of her perfume, a whisper of lavender and spice, seemed to linger in the air. He pictured her, laughing, and he felt a pang, a sharp ache, deep in his chest. A sudden urge to just… do something, anything, surged through him. He decided he would, even if it was just to honor her memory.

Emotion: reflective

Cluster: Nostalgia / Reflection
PC1 (Valence): -1.27 Negative
PC2 (Disposition): -1.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.650
L0.547
如今0.402
H0.353
own0.319
Suppressed:
la-0.687
de-0.659
l-0.388
/-0.324
--0.320