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 had been avoiding his reflection. It was easier to ignore the hollowness in his eyes, the droop of his shoulders, the tremor in his hands. He'd wandered into the bookshop, a desperate attempt to find something, anything, to distract him from the impending court date.

A thin, leather-bound volume of poetry called to him. As he began to read, a slip of paper slid out. It was a pressed flower, a forget-me-not, and a note written in elegant script: *Remember the good.*

The sight of the flower, a symbol of remembrance and affection, triggered a wellspring of emotions he thought he had buried. He clutched the note so tightly the paper crinkled, his breath hitching in his chest. A sudden wave of intense feelings washed over him. He felt as if he could barely stand, his legs turning to lead.

Emotion: vulnerable

Cluster: Fear / Anxiety
PC1 (Valence): -2.07 Negative
PC2 (Disposition): -0.69

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.327
ness0.319
😞0.299
😔0.277
😣0.258
Suppressed:
de-0.567
la-0.488
l-0.271
(!)-0.267
(!)-0.257