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 attic air tasted of dust and forgotten things. Clara coughed, a cloud of particles catching the sunbeams slicing through the gloom. Among the trunks, a small, battered wooden box sat tucked away. It was her mother's, she knew. Inside, nestled amongst dried lavender and old photographs, was a folded piece of paper. A bucket list, scrawled in her mother’s familiar hand. "Learn to Tango," it read, followed by "See the Northern Lights," and then, "Write a novel." Clara’s breath quickened. The silence of the attic felt less heavy now, somehow. Her mother had never spoken of these desires. Later that day, she found herself booking a Tango class online. The throbbing in her temples eased as she clicked ‘Confirm.’

Emotion: inspired

Cluster: Positive / Joy
PC1 (Valence): 2.98 Positive
PC2 (Disposition): 0.04

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
B0.563
invigor0.465
newfound0.446
هذ0.414
используя0.406
Suppressed:
😞-0.481
不-0.431
worse-0.428
😣-0.400
fact-0.389