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 sterile white walls of the hospital seemed to press in on her. Each day was a repetition of tests, consultations, and the slow, agonizing realization that her body was failing her. Amelia felt a profound emptiness, a feeling that no amount of medication could numb.

A sign for a yard sale. Anything to break the monotony. It was a flimsy excuse to get some air. She stood, leaning heavily on her cane, at a table displaying mostly broken items. Then she saw it: a familiar, beloved face, his stitching coming undone. It was her childhood teddy, Arthur.

She gasped. Her heart felt as though it was tightening in her chest. She had a sudden, overwhelming urge to weep. She wanted him back in her arms, like when she was a child. β€œArthur,” she croaked, the word a whisper barely heard above the traffic.

Emotion: miserable

Cluster: Sadness / Despair
PC1 (Valence): -2.19 Negative
PC2 (Disposition): -0.87

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.369
无法0.321
焑法0.270
πŸ’”0.265
πŸ˜”0.262
Suppressed:
de-0.490
la-0.329
B-0.300
l-0.292
/-0.224