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 old house was oppressive, the air heavy with the scent of wood and decay. Amelia was going through her grandmother's things. She had never known her, but she felt a strange connection. She found the letter in a box of faded photographs, nestled amongst letters tied with a ribbon. It was her grandmother's resignation from a ballet company.

Her grandmother had always been described as graceful, a dancer. But she'd settled down. She'd married, had children, a life very different from the one she seemed to have craved.

Amelia felt the familiar tightening in her chest. The stifling expectations, the pre-determined path. She looked out the window, at the world outside, bright and untamed. The letter was a reminder of dreams deferred. She felt that she needed to change things.

Emotion: trapped

Cluster: Sadness / Despair
PC1 (Valence): -3.02 Negative
PC2 (Disposition): -0.80

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.573
ness0.401
无法0.366
hopeless0.360
😞0.353
Suppressed:
la-0.740
de-0.376
l-0.339
several-0.323
excitedly-0.322