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).

The metallic tang of fear coated his tongue. He tried to take a deep breath, but the air felt thin and inadequate. He had a million things to do, and now he was stuck. He could already envision the emails, the phone calls, the missed opportunities.

He paced the confines of the elevator, a caged animal. Every mechanical noise sent a jolt of alarm through him. He rubbed his temples, a futile attempt to soothe the pounding headache that had begun to blossom behind his eyes.

"Is there any way to...?" he began, his voice barely a whisper, then stopped, realizing the futility of the question. He kicked at the wall, the sound echoing in the enclosed space. The building's maintenance was so unbelievably shoddy.

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