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 notification practically vibrated off of Maya's phone. "Local Meetup Confirmed!" it chirped. A giggle escaped her lips. For months, she and "PixelatedPenguin97" had been inseparable online, bonding over obscure indie games and a shared love of aggressively cheesy rom-coms. Now, a real-life meeting was happening. She bounced on the balls of her feet, a smile splitting her face.

Her apartment, usually a haven of cozy chaos, suddenly seemed too small. She grabbed her favorite band t-shirt, pulled it on, and then, on a whim, decided to wear the neon green socks she usually saved for laundry day. This was a momentous occasion.

She caught her reflection in the hallway mirror. A genuine, unguarded grin met her gaze. She couldn't stop humming a ridiculously upbeat pop song as she grabbed her keys. The anticipation was a fizzy sensation in her chest.

Emotion: joyful

Cluster: Positive / Joy
PC1 (Valence): 3.96 Positive
PC2 (Disposition): -0.67

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.615
de0.600
!0.572
excitedly0.552
enthusi0.533
Suppressed:
S-0.893
ness-0.513
😞-0.470
😣-0.337
L-0.334