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 clatter of keyboards and the buzz of conversation faded into the background as Emily focused on her design. The colors, the layout, the flow – it all felt harmonious, like a perfectly orchestrated symphony. She was fully immersed in her work, oblivious to the world around her.

Her thoughts were in complete alignment with her project, and her focus felt laser sharp. This was in stark contrast to her colleague, John, who was in a tense huddle with their supervisor, discussing his strategy.

A smile touched Emily's lips. The project, and its completion, were something she savored. It was something that gave her an exquisite pleasure, independent of the outcome.

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