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).

He held the list, the paper thin and worn, his father's final wishes. "Learn to play the guitar." It seemed a cruel joke, given his father's notoriously tone-deaf singing. A sudden urge seized him: he wanted to do it, truly do it.

His limbs felt lighter, his breath deeper. He immediately went online and ordered a beginner's guitar. The clicking of the mouse resonated in the quiet room. He felt a spark of something new, something that chased away the heaviness that had clung to him for so long.

He imagined his father's smile, a quiet, amused smile. He would learn. He had to. He would play until the sun went down, and when the sun went down he would start all over again.

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
πŸ˜₯-0.385