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).

"Alex, we need you in the backstroke." The swimming coach's words hit Alex like a tidal wave. He had spent years specializing in freestyle, the only stroke he'd ever truly known. He felt a surge of discombobulation.

His gaze flickered around the pool, the familiar blue water now seeming to mock him. He swallowed hard, a lump forming in his throat. He saw the other swimmers, their bodies sleek and efficient, preparing for their sets. He felt a wave of doubt crash over him.

He cleared his throat. "But… backstroke?" The question came out as a squeak. His entire body felt leaden, his usual confidence completely drained. It felt like he'd been dropped into a different world.

Emotion: astonished

Cluster: Surprise / Confusion
PC1 (Valence): -0.84 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
que0.448
de0.350
la0.323
l0.307
incred0.266
Suppressed:
own-0.289
cheerfully-0.231
enjoyed-0.224
开心-0.221
ี่ยว-0.219