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 walked into the workshop, the familiar scent of metal and oil clinging to the air. The place was empty. The door was unlocked, a clear indication the workshop was not in use. A note was taped to the workbench: "Moving on. Cheers." He reread it a dozen times, his jaw clenching.

He picked up a wrench, turning it over in his hand, remembering all he had learned. The wrench felt heavy, useless. He felt a sudden, crushing weight on his shoulders.

He felt betrayed. A burning sensation started behind his eyes. He had assumed he would have a farewell, at the very least. He had expected more. He felt suddenly alone.

Emotion: reflective

Cluster: Nostalgia / Reflection
PC1 (Valence): -1.27 Negative
PC2 (Disposition): -1.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.650
L0.547
如今0.402
H0.353
own0.319
Suppressed:
la-0.687
de-0.659
l-0.388
/-0.324
--0.320