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 weight of the violin against her shoulder, the feel of the bow gliding across the strings, the music filling the concert hall – all of this was home to Isabella. As the applause thundered around her, she felt a profound sense of… completeness.

She discovered the letter while sorting through some of her grandmother’s old letters and photographs. The handwriting, spidery and delicate, was unmistakable. It was a draft resignation letter from her grandmother’s teaching position, dated 1944.

Isabella read the letter, filled with wistful phrases. Her grandmother had wanted to pursue a career in music. A shadow of disappointment. Isabella felt a twinge of… something. She took a deep breath. She glanced around the stage, the other musicians, the audience. She picked up her bow and began again. She knew she was living her grandmother’s dream.

Emotion: fulfilled

Cluster: Positive / Joy
PC1 (Valence): 3.22 Positive
PC2 (Disposition): -1.74

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.819
மகி0.415
happy0.397
grateful0.389
de0.385
Suppressed:
S-0.599
worse-0.465
C-0.460
either-0.436
😣-0.426