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 box of old photographs lay spread across the living room floor, each image a tiny portal to the past. Sarah stared at them, the faces of her and Emily, young, carefree, a kaleidoscope of shared adventures. The ache in her chest seemed to grow with each photo. Emily, moving to New York. The city. A world away.

She traced her finger over a picture of them laughing at a concert, the music loud, the night alive. She’d been so happy then. Now, a deep, wearying sadness settled upon her, the apartment suddenly feeling small, suffocating.

She picked up a photo of them at the beach, salty air, the sun, the laughter. The future felt like a blank page, the story of their friendship unfinished. The room, filled with memories, also felt empty.

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