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 old house felt colder than usual. Sarah shivered, despite the fire crackling in the hearth. She was looking at the box her mother had left behind; photographs mostly. She'd always understood her mother to be reserved, even distant. Her own childhood, she now realised, was filled with a sense of lacking, a void she could never quite put her finger on.

One particular photograph caught her eye: her mother, young and radiant, dancing in a crowded ballroom. Beside her, a man; his expression was pure devotion. She never spoke of any such people, she had always seemed to live alone.

Sarah felt a stab of something she couldn't name: betrayal, maybe? Or was it sadness? The woman in the picture looked so happy, so free. She looked nothing like the mother she'd known. The secret that had been kept from her.

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