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 critique, mercilessly honest, lay crumpled on the floor. "A tragedy of textures, a culinary insult," read the review. For Daniel, the master of his domain, the words were a personal affront. He couldn't seem to breathe right. His vision swam.

He slammed his fist on the counter, the force a release of his frustration. He left the restaurant, the empty street offering no comfort. He wandered aimlessly, the city's hustle and bustle a discordant symphony to his despair.

He sat on a bench, head in his hands, feeling utterly lost. The sun set, casting long shadows that mirrored the deep emptiness inside him. He felt like his purpose was gone.

Emotion: worthless

Cluster: Sadness / Despair
PC1 (Valence): -2.10 Negative
PC2 (Disposition): -0.15

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
L0.427
S0.405
😞0.386
ness0.339
缺乏0.326
Suppressed:
de-0.320
(!)-0.286
随后-0.268
😍-0.262
tc-0.255