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 cold air outside had made Alice’s cheeks rosy. She had just found a fantastic bargain on a rare book, a find that had her practically vibrating with delight. Today felt like a day of unexpected treasures.

She wandered through the shop, her eyes scanning the various items. Suddenly, a ring, sparkling in a display case. It was the exact style of the ring her grandmother had lost during a move.

A surge of emotion, a mix of excitement and disbelief, washed over her. She practically bounced over to the counter. “Oh, could I take a look at that ring, please?" she asked, her voice carrying an infectious enthusiasm. The shop owner, a man with a tired face, simply pointed to the locked case.

Emotion: cheerful

Cluster: Positive / Joy
PC1 (Valence): 4.13 Positive
PC2 (Disposition): -1.19

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.962
de0.597
!0.590
excitedly0.589
!".0.568
Suppressed:
S-1.213
L-0.693
ness-0.492
😞-0.425
বারবার-0.422