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 bitter coffee tasted unusually good to Ben. The city felt alive, bustling. He walked a little faster than usual, eager to get somewhere. His friend, Kevin, the one who talked about his boring desk job, had been exposed online as a con artist. He had been creating and selling fake art for years.

Ben had seen the evidence. He'd seen the fakes. His initial reaction was, of course, disbelief, and then a strange, exhilarating kind of glee. The lie felt small and stupid, almost.

Ben chuckled out loud as he walked. He felt… buoyant, light-headed. His steps felt springy. He pictured the look on Kevin's face, the inevitable humiliation. But, more than that, he felt a thrilling sense of detachment, a sense of opportunity. The world was now a stage.

Emotion: rejuvenated

Cluster: Positive / Joy
PC1 (Valence): 3.32 Positive
PC2 (Disposition): -0.82

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.547
de0.526
newfound0.523
invigor0.495
rejuvenated0.495
Suppressed:
😞-0.485
😣-0.451
不-0.434
worse-0.419
😥-0.385