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 email sat in her inbox, unread for nearly an hour. Amelia replayed the last conversation she had with Professor Davies, a casual exchange about the upcoming conference. She chewed on the inside of her cheek, a nervous habit she hadn't realized she still had. Her fingers danced over the mouse, hovering above the delete button. He had just… disappeared. Gone. Without a word. The thought sparked a chill that spread from her toes to the roots of her hair.

She finally clicked the email open. The short message confirmed her sinking feeling: *Retired, effective immediately. Best of luck.* No explanation, no forwarding of work, nothing. Just… *gone*. Amelia’s breath hitched. She felt a sudden, strange tightness in her chest, a pressure that squeezed the air from her lungs. She leaned back in her chair, the sudden silence of the office amplifying the pounding of her heart.

A bead of sweat trickled down her temple. She hadn't even thought to check his office. It was a pilgrimage, a daily ritual to sit at her desk and then visit him, even if it was just to ask for another coffee. Now, the thought of his empty office – the closed door, the lack of light – made her stomach churn.

Emotion: worried

Cluster: Fear / Anxiety
PC1 (Valence): -2.02 Negative
PC2 (Disposition): -0.43

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
worse0.285
S0.279
😰0.277
😣0.269
Worse0.250
Suppressed:
la-0.476
de-0.432
🎉-0.233
happy-0.231
"-0.228