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 silence in the kitchen was thick as Ms. Alvarez, my neighbor, stood across from me, sipping her tea. The tree outside, a magnificent maple, cast dancing shadows on the walls. I found myself focusing intently on the way the steam curled from her cup.

She smiled, a forced expression that didn't reach her eyes. Her tone, usually warm, was clipped and business-like.

"The roots are… troublesome," she said, her gaze drifting towards the window.

I felt a subtle shift in the air, a tightening in my chest. The idea of the tree being gone felt wrong, a severance of something vital. I noticed a slight tremor in my own hand.

"I need to consider all angles," I said, hoping my voice sounded steady. The maple's leaves whispered secrets in the breeze, and I suddenly felt a pressing need to run outside, to touch its bark, to reassure myself it was still there.

Emotion: uneasy

Cluster: Fear / Anxiety
PC1 (Valence): -1.80 Negative
PC2 (Disposition): -0.43

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😰0.240
πŸ˜–0.231
😣0.231
느껴0.226
worse0.222
Suppressed:
de-0.474
la-0.465
happy-0.261
a-0.255
ecstatic-0.228