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 phone rang, and the caller ID showed Mrs. Davies’s number. Mark hesitated. It always felt like a threat. He picked up, bracing himself.

“Mark, dear, could we discuss the willow?” Her voice was all sweetness and steel.

The willow. Its weeping branches had always felt like a comfort, a curtain around his small space. Mark’s shoulders slumped. He’d barely left the house in weeks.

“Yes, Mrs. Davies,” he said, his voice a low monotone. He felt his pulse quicken.

She outlined her concerns – the roots, the potential damage to her fence. Mark listened, a cold dread creeping through him. He agreed to whatever she suggested, anything to avoid further interaction, further exposure. When he hung up, he felt a profound emptiness.

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