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 world felt muted, a dull grey. Liam found himself moving through life on autopilot. Getting out of bed was an ordeal. He skipped meals, lost his appetite. Simple tasks felt like monumental efforts. He’d lost interest in his hobbies, the things that once brought him joy. His friends kept calling, but he couldn't find the energy to answer.

The therapist’s office, a cozy space with overstuffed armchairs, should have been comforting. Instead, the sight of the woman behind the desk sent a wave of nausea washing over him. It was Susan, the girl who’d made his life a misery in high school. The queen bee, who had always reveled in his suffering.

"Liam," she greeted him with a polite smile, "Please, have a seat." The smile felt brittle, fake.

"I understand you've been having a difficult time," she said. The phrase hung in the air, heavy with unspoken history. He felt the familiar sting of shame, the urge to retreat.

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