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 scent of pine and wood smoke always made Charlotte feel at ease as she entered The Lakeside Lodge. It was her place of solace, a place where she could sit by the fire and escape the everyday hustle of the city. She loved the atmosphere, and the anticipation of a warm meal and a good book filled her with excitement.

The CLOSED sign was like a slap in the face. She felt a moment of sadness. Her happy place? Gone? She felt a small wave of disappointment.

She took a deep breath, and her shoulders went back. She would not let this stop her. There was a new hiking trail she'd heard about, and a new book she'd been itching to start. The sun was shining, inviting her to start a new adventure. She started walking, her heart filled with new possibilities.

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
๐Ÿคฉ0.515
Suppressed:
S-0.891
๐Ÿ˜ž-0.504
L-0.486
ness-0.478
ใใ‚Œใงใ‚‚-0.432