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 rhythmic clack of the train tracks lulled me into a strange state of stillness. My mind, usually a whirlwind of activity, felt strangely blank. I gazed at the passing scenery, the blurred fields and trees, and realised I had to get this project finished. My fingers drummed on the worn leather of my bag. This book... It held everything.

A woman sat across from me, her eyes locked on a laptop. She was writing. Suddenly, she looked up, her gaze meeting mine. Her eyes widened, a flicker of surprise crossing her face.

"Is that...?" she began, gesturing towards my bag. I felt a chill, a tightening in my throat. I already knew. The way she was looking at me, the book open on her lap – the cover of my unfinished novel, my heart sank.

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