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 courtroom was vast and echoing, the polished floor reflecting the harsh fluorescent lights. I couldn't seem to get a proper breath. My chest felt constricted, as if an invisible hand were squeezing the air from my lungs. The judge’s pronouncements were a distant drone, each word a hammer blow to my composure.

My sister, Sarah, inherited the family business, a successful chain of restaurants. Her face was lit up; she was the golden child. I, on the other hand, received a box of my father's old tools and a half-finished manuscript, his life's work. I wanted to run, to escape this stifling place.

"It’s just… a lot to process," I mumbled when she started congratulating me.

Emotion: overwhelmed

Cluster: Fear / Anxiety
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -0.94

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.385
😫0.332
😖0.319
😣0.319
😩0.314
Suppressed:
la-0.576
de-0.496
a-0.297
happy-0.290
l-0.264