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).

"So, the interview went well," the HR manager had said, a week ago. “Both candidates have excellent portfolios.” Now, staring at the message on his phone, he felt a strange lightness, like the world had somehow lost a layer of gravity. Him and David? Him and David, who barely knew how to use PowerPoint?

He gripped the phone tighter, the smooth glass cool against his sweaty palm. A knot formed in his stomach. He wasn’t sure if it was a good feeling. It felt more like a seismic tremor in his gut. A low hum filled his ears, a constant reminder of the absurdity of it all.

He reached for his coffee, then hesitated, his hand frozen mid-air. It felt wrong to drink it now. Everything felt wrong.

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