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).

Sunlight streamed through the kitchen window, warming Amelia's face as she listened to Mrs. Gable's proposal. The request, to utilize a strip of land near the back fence for a flower garden, seemed almost too good to be true.

"Of course, Mrs. Gable! Please, take as much space as you need," Amelia responded, her hands clasped tightly. Her fingers tapped a quick rhythm against the countertop, an involuntary movement. She found herself imagining the vibrant blooms already, picturing the splashes of color against the green grass.

"Are you sure, dear? I wouldn't want to impose." Mrs. Gable hesitated.

"Not at all! I'll even help you fence it off, if you'd like. I've always wanted more color in the yard." A grin split Amelia's face, the corners of her eyes crinkling.

Emotion: eager

Cluster: Positive / Joy
PC1 (Valence): 2.98 Positive
PC2 (Disposition): 0.02

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.659
excitedly0.436
!0.389
!"0.386
ecstatic0.383
Suppressed:
S-0.816
ness-0.441
😞-0.387
L-0.379
P-0.376