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).

"No way," Emily whispered, rereading the grade on her phone. A "D" on the calculus exam. She’d spent the weeks leading up to the test feeling distracted, easily led astray by the allure of distractions. Now, she felt a creeping sense of panic.

She shoved the phone in her pocket and walked towards the library, the familiar scent of old books doing nothing to calm her nerves. She picked up a novel, intending to study. But the words blurred before her eyes, and she closed the book, the weight of the upcoming exams heavy on her shoulders. She should probably study. Instead, she found herself staring at the other students, their intense expressions making her feel isolated.

Emotion: lazy

Cluster: Fatigue / Lethargy
PC1 (Valence): -0.37 Negative
PC2 (Disposition): -0.02

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😴0.458
snooze0.425
惰0.401
Lazy0.392
躺0.386
Suppressed:
able-0.371
感激-0.289
because-0.274
因-0.273
কারণ-0.272