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 morning sun streamed through the kitchen window, warming Amelia's face as she hummed, flipping pancakes. Mr. Henderson’s house next door had been a drab, beige box for as long as she could remember. Now, a massive dumpster sat in his driveway, and a crew of burly men in hard hats were already swarming the place. Finally! Something interesting to look at besides the squirrels.

She poured herself a cup of coffee and sauntered onto the back porch, her hand resting on the smooth wood of the railing. The rhythmic hammering already had a pleasant energy about it. This would be a welcome change of pace, something to shake up the monotony of her quiet life.

“Morning, sunshine!” she called out to the lead carpenter, a man with kind eyes and a welcoming smile. He waved back, and she grinned, the corners of her mouth twitching upwards.

Emotion: elated

Cluster: Positive / Joy
PC1 (Valence): 3.38 Positive
PC2 (Disposition): -0.23

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.724
!0.596
ecstatic0.510
🤩0.510
overjoyed0.504
Suppressed:
S-0.657
😞-0.497
ness-0.407
😔-0.380
😓-0.377