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 found it nestled beneath the old oak tree in his backyard, a plastic container that looked as if it had been buried yesterday. The label, scrawled with his own familiar penmanship, boasted of a future he had no recollection of hoping for. A knot tightened in his stomach.

He sat on the damp grass, the capsule resting on his lap. He felt a profound sense of wrongness, as though something vital had been disconnected. He tried to think, to grasp a memory, but only found an empty void.

He cautiously lifted the lid. Photos of faces he didn't recognize, items that held no personal meaning, and a journal filled with a script he didn’t recognize filled the container. His mouth felt dry. The images from the pictures seemed to be mocking him.

Emotion: disturbed

Cluster: Fear / Anxiety
PC1 (Valence): -2.97 Negative
PC2 (Disposition): -0.60

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.318
ness0.309
无法0.300
😞0.297
worse0.284
Suppressed:
de-0.829
la-0.666
a-0.392
B-0.301
"-0.290