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).

Thomas sank into the worn armchair, the album heavy in his lap. The late afternoon light cast long shadows across the living room, painting the familiar space in a melancholic hue. He hadn't touched this album in years, not since his mother's passing. Now, with the house feeling empty, he needed…something.

The first few pages showed the idyllic life: his parents, smiling, holding him as a baby, the family dog, all the expected milestones. But then he found it: a picture of his father, looking younger than Thomas had ever seen him, leaning against a vintage car, with a woman he didn't recognise. She was wearing a dress that was all wrong for the time period, her arm casually resting on his father's shoulder.

He felt a sudden, sharp pang, a hollowness in his stomach. His mother had always spoken of a youthful romance, a boy she'd loved before meeting his father. But this… this felt different. It was a secret, a hidden piece of the puzzle that changed the whole picture. He traced the woman's face with his finger, a knot forming in his throat.

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