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).

Elias, hunched over his laptop, battling writer's block, felt utterly defeated. He hadn't written a decent lyric in months. His fingers hovered over the keys, paralyzed. Suddenly, a familiar melody, played on a piano, drifted from the open window. It was โ€œForgotten Dreams,โ€ a song heโ€™d abandoned years ago.

He threw his head back, and laughed out loud. The tune was being performed with a certain elegance. His shoulders straightened, and he felt a burst of warmth spread through his chest. He felt like he could stay awake all night. He walked over to the desk, and opened a new document, ready to write.

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
๐ŸŽ‰0.434
invigor0.416
enthusi0.395
Suppressed:
๐Ÿ˜ž-0.524
๐Ÿ˜”-0.428
๐Ÿ˜ข-0.404
๐Ÿ˜ฅ-0.393
que-0.391