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 wind changed, and with it, the scent. The smell of fresh-cut wood, now mingled with the sharp bite of uncertainty. It filled her nostrils, a constant reminder. She tried to locate herself in the physical space, but everything seemed wrong. The furniture, the walls, the very air she breathed.

She felt a deep disconnection, a profound sense of isolation. The fence. Of course, the fence. It was finished now, a solid, unyielding barrier. The world beyond felt distant and blurred.

She walked towards the window, her steps slow and uncertain. The fence stood between her and the outside world, a boundary she didn't understand. She felt a profound sense of helplessness, a feeling that she had lost the path and couldn't find her way back.

Emotion: disoriented

Cluster: Surprise / Confusion
PC1 (Valence): -2.65 Negative
PC2 (Disposition): -1.34

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
unfamiliar0.340
ness0.321
失去了0.316
не0.315
будто0.311
Suppressed:
B-0.474
la-0.419
K-0.409
de-0.309
brag-0.302