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 paper felt thin and fragile, but the words held a potent energy. Evelyn gazed at the list, her husband, Mark’s secret ambitions laid bare. "Learn to fly a plane." "Visit every national park." "Write a song." A sensation, almost electric, vibrated through her.

She picked up the phone and started searching for flight schools in the area. The thought of Mark behind the controls, a wide smile on his face, filled her with a sense of excitement.

She got a guitar from the attic, a forgotten instrument. The strings felt cold beneath her fingers. She started to pluck tentatively, the beginning of a melody forming. This was going to be fun.

Emotion: invigorated

Cluster: Positive / Joy
PC1 (Valence): 3.28 Positive
PC2 (Disposition): 0.00

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
!0.452
(!)0.439
πŸŽ‰0.434
invigor0.416
enthusi0.395
Suppressed:
😞-0.524
πŸ˜”-0.428
😒-0.404
πŸ˜₯-0.393
que-0.391