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 new spreadsheet was complex, and the data felt… manageable. For weeks, Amelia had felt a dull ache behind her eyes, but today, that pain was gone. The sunlight spilling through the blinds warmed her face, and a slight tremor of excitement ran through her veins.

During a break, a new intern mentioned their love for canoeing.

“Canoeing?” Amelia repeated, a warmth spreading through her chest. “I used to canoe at Camp Whispering Pines! Do you know it?”

The intern’s jaw dropped. “No way! My parents met there! My dad was a lifeguard.”

Amelia laughed. “I remember the lifeguard! He had the craziest tan lines.”

“And you were there?” The intern seemed overjoyed.

“Absolutely! I was a terrible swimmer, but I loved the campfires.” A happy sigh escaped Amelia’s lips.

Emotion: optimistic

Cluster: Positive / Joy
PC1 (Valence): 4.44 Positive
PC2 (Disposition): -0.49

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.856
la0.688
!0.601
enthusi0.530
🤩0.515
Suppressed:
S-0.891
😞-0.504
L-0.486
ness-0.478
それでも-0.432