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).

"Well, this is… interesting," said Silas, staring at the paperwork. His sister, Iris, beamed. They had just finished reading their parents' will. Iris had been given their parents’ bustling restaurant, a place filled with laughter and the aroma of cooking. Silas was left with a collection of their mother’s unpublished poetry.

He wandered to the window, the sun a pale disc in the sky. He found himself unable to work. The thought of poems, of words and emotions, filled him with a weariness that seemed to emanate from his very bones. He felt the emptiness pressing.

He opened one of the books, the pages brittle with age. He read a few lines, his eyes tracking the words. He closed the book, the feeling of despair washing over him again. He was in a cold cavern, and it seemed as though he could not escape.

Emotion: depressed

Cluster: Sadness / Despair
PC1 (Valence): -2.06 Negative
PC2 (Disposition): -1.08

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.417
每天0.353
无法0.330
own0.311
😔0.309
Suppressed:
de-0.492
B-0.393
P-0.352
l-0.310
!-0.307