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 weight of the laundry basket felt insignificant compared to the crushing burden on David's shoulders. The scent of fabric softener did little to mask the lingering metallic tang of the day's discovery. The hidden account. The betrayal. He tossed the damp clothes into the dryer, the rhythmic tumbling a counterpoint to the chaotic thoughts swirling in his head. While doing a budget, he'd noticed unusual transactions. He was so stunned, he'd had to sit down.

He leaned against the humming appliance, the vibration a dull thrum against his back. He felt a sense of profound disorientation, as if the ground had fallen out from under him. He rubbed his temples, trying to ward off the throbbing ache that was rapidly spreading across his forehead. He was in shock. He wanted to run.

Emotion: overwhelmed

Cluster: Fear / Anxiety
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -0.94

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.385
😫0.332
😖0.319
😣0.319
😩0.314
Suppressed:
la-0.576
de-0.496
a-0.297
happy-0.290
l-0.264