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 gentle clink of silverware and the hushed tones always calmed Eleanor as she approached The Cozy Corner Cafe. It was her haven, a place where she could enjoy a quiet afternoon tea and catch up with her thoughts. She felt a peaceful anticipation build as she neared the entrance.

The CLOSED sign seemed to loom over her, a dark cloud hanging in the sun. She felt a sharp pang of disappointment, a momentary loss. This wasn't supposed to happen.

She straightened her shoulders. Well, that was that. No use in dwelling on it! She could explore a new place! She saw an ad for a new tea house the other day. It was time for an adventure. A new chapter. She smiled and started walking.

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