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’d almost broken the steering wheel on the drive to pick up Liam. The radio blared a cheerful pop song, but it only amplified the pressure building in his chest. Mr. Henderson. The name, echoing in the school's digital notification, was like a physical blow. Of all the teachers…

He pulled into the drop-off lane, his hand tight on the gearshift. Liam hopped in, chattering about art class, oblivious. He forced a smile.

“How was your day, buddy?” he asked, trying to sound normal.

Liam’s face lit up. “It was great! Mr. Henderson is really cool.”

The words were a physical jab. He managed a strained laugh. "That's…fantastic." He drove home in silence, his fingers drumming an erratic rhythm against his thigh. He was going to have a talk with Liam later tonight, a long talk.

Emotion: mad

Cluster: Anger / Hostility
PC1 (Valence): -1.94 Negative
PC2 (Disposition): 1.26

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.448
骂0.344
est0.340
🤬0.336
aggravated0.332
Suppressed:
a-0.495
de-0.485
la-0.387
H-0.344
latter-0.290