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 elevator was suffocating. Michael watched the numbers flash by, each floor a step further away from the solace of his apartment. The feeling of desolation clung to him like a second skin.

“You ever go to Camp Skylark?” asked his new coworker, John, as they exited the elevator.

Michael stopped dead in his tracks. Skylark. The very name felt like a betrayal. He was supposed to feel joy when hearing it, not the way his chest now constricted.

“Yeah,” Michael managed, his voice a low growl. He avoided eye contact, desperate to escape the inevitable small talk.

John launched into a story about a canoe trip, his voice ringing with happiness. Michael wanted to scream, to lash out, but instead, he simply walked on, the emptiness inside him growing with each step.

Emotion: heartbroken

Cluster: Sadness / Despair
PC1 (Valence): -2.60 Negative
PC2 (Disposition): -1.10

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.433
S0.385
😔0.359
haunting0.349
😢0.342
Suppressed:
de-0.503
la-0.455
B-0.352
🤓-0.332
!-0.299